Club des Responsables d’Infrastructures et de Production IT Infrastructure and Operations

IT Infrastructure & Operations Management Best Practices

MAINFRAME IBM zSeries Mainframe Cost Control te PAPER

Authors Laurent Buscaylet, Frédéric Didier Bernard Dietisheim, Bruno Koch, Fabrice Vallet

Sponsored by September 2012 Wh i PAGE 2 4. CONL 3. ZPLATFORM COSTONTROL 2. zPlatform O 1. “z”PositioningandStrategy withinaCompan y INTRODUC T CRiP TABLE OFFIGURES 5.  of AND A APPENDIX -GLOSSARY OFTECH NIC 3.2 Infrastructure Cost Reduction 3.1 Controling Software Cost 2.2 Mainframe software vendors (ISV’s) analysis &positioning 2.1 Technical terms ofreference 1.4 Evolution Strategies for System z 1.3 System z-Banking/Finance/Insurance vs. General Industry Sectors 1.2  1.1 Role oftheMainframe intheeyes ofCRIP’s Members 3.3 Controlling operational costs 2.3 Survey Results able The Position ofSystem zvs. Distributed Systems (Windows, UNIX,Linux) 3.2.2 Useofspecialtyengines:zIIP,zAAP,IFL 3.2.1 Grouping andSharingInfrastructure 3.1.4 MLC: Controlling thebillinglevel 3.1.3 ISVcontractual arrangements: licensing modes&billing 3.1.2 ISVnegotiationstrategies: holistic/generic orpiecemeal/ad-hoc 3.1.1 IBMContractual Arrangements 2.2.2 TheSecondary Suppliers (ISV’s) 2.2.1 TheMainSuppliers (ISV’s) 2.1.6 Business Continuity 19 2.1.5 T 2.1.4 Backups 2.1.3 Storage 2.1.2 SAN&Networks 2.1.1 Servers 3.3.2 CapacityManagement,tools &methods 3.3.1 Optimization,performance &technical/application component quality contents CRONYMS USEDINTHIDOCUMENT USION 3.1.4.5 Other Cost Saving Options 3.1.4.5 OtherCostSaving feedback andrecommendations3.1.4.4 Survey 3.1.4.3 Themethods andcontrols for controlling thesoftware invoice 3.1.4.2 Theproblem of smoothing (consolidating) peakloads 3.1.4.1 TheIBMbillingmethod TION echnical Architecture &Organization verview AL TERMS 28 11 67 73 72 68 50 28 20 11 55 54 50 38 35 31 28 22 16 15 13 12 11 58 55 48 43 42 41 38 22 6 5 9 7 6 6

PAGE 3 PAGE 4 Foreword Laurent Maneville, Global Business Manager, zCost Management Michael Moss, ManagingDirector ofValue-4IT English translation : Jérôme Viguier Hao ungHoi Patrice Rosier Pierre Fenaroli François-Xavier Ducreux Marie-Françoise Diver Thierry David Hervé Combey Romain Capron Sandra Belinguier Alexandre Babin Contributors to theMainframe’s workgroup : Editor : RenaudBonnet,CRiP Fabrice Vallet Bruno Koch Bernard Dietisheim Frédéric Didier Laurent Buscaylet Authors ofthisWhite Paper : Please enjoy! insurance. and bank industry, as such business of lines various from group users French a CRIP associationMainframecustomers)(independent of by the among written been has paper white This After leading this group and writing up this white paper global other companies. to applicable and topical very still are (cost practices control, capacity management) promoted context, in this paper economic global today’s In recognized byeveryone. toall haveof one’sat elementsdisposal accepted and to federate all parties, to obtain a consensus and most effectiveimplementation. architecture …), and also final users, is necessary one’s to its (production, company make the of entities variousInvolving and and optimize manage, mid-term. in way infrastructure centraloptimum evolve an in to the lever remainsvital management Capacity significant reach to objectives inavery shortterm. allow and users Mainframe of and tools are many; they are applicable to the majority Solutions axis. work critical a been has control Cost the majorconcerns for theseusers. of one represented already 2010 of beginning the at T ools and processesand enableools Air France Air France CNP La BanquePostale i-BP Natixis i-BP La BanquePostale Matmut La BanquePostale La BanquePostale PSA PEUGEOT-CITROËN ex-GCE Tech (Groupe Caisse d’Epargne), work group’s Leader Renault ex-Crédit Foncier deFrance PSA PEUGEOT-CITROËN T opics chosen opics promotion oftheMainframe ecosystem. the to participate therefore,to and level, worldwide a on works quality high such publishing help to able be to this and sponsor their Paper. of be White to proudareWe translation the for company our chose CRIP your company. efficient has morepromoteprojectsin you help will and topics, these on be to Management you that allow convinced will zCost are information this we CRIP, since paper the this translated with line In economic already madesustainable. unprecedented the investments and Mainframe making whilechallenges face to our allow customers to is purpose Our servers. Mainframe on management capacity and control cost in specialized arewe 2012; groupin zCost the joined I 2010-2011, in zCost Services ManagingDirector zCost Management Founder Jacky Hofbauer Bruno Koch SzLC, zNALC,areincrediblycomplex;ULC,IWP, AWLC, IPLA, contractsSoftwareand licensesthat is Mainframe IBM the of aspects unpleasant the of one that admit Let’s solution would bethemost cost efficient? which al, et operated, up, backed administrated, be to need which servers x86 many on installeddeployed, is application x86 an once Indeed, mistake.classic one) a servers; (or is few a infrastructure, on deployment to global compared often a on typically Mainframe deployment, IBM Furthermore, ease. operational administration and of terms in counterpart donot great the see who managers financial confuse for Software investment, all above figures, UNIX cost high alleged The astonish would nature. by global is environment their and colleagues!) rate utilization administratorsmostlikely(and Windowsinfuriate their their and consolidated hyper are servers Series often z criticism; this in is distortion cognitive and many. irrational of by sort a platform is high There too considered Mainframe are prices IBM distrusted; the However, decreasing, butstill huge. advanced Distributed Systems platform; admittedly it is gap between virtualization on System z and on the most features.same” “the providealmost they so; do to managing really without Mainframe the this to by trying it. and equal it inspired reinventing it, being imitating keep model, keep Systems Distributed business. big many so organizations why still use the IBM Mainframe for their core explain qualities great these of All features. Recovery, clustering Disaster and remote decades.replication are forobvious Mainframe IBM the on running reliably been has Java acronyms. VTL and SAN the of tape and used toolswerevirtualization Sharedevenused before inventionthe been more. have or networks years service twentystorage for tasks have daily management been capacity and balancing workload automation, Operations time. long a forcharacteristics major its of two been have sharing infrastructure and Consolidation servers. x86 on than earlier years thirty And yet virtualization was possible on Mainframe at least believe it,can you? cannot trendy. You very still is but fifties its is in almost Mainframe IBM the is 1964, April it in introduced been and “dinosaur” year. each modern modernized increasingly most becoming the is it Firstly Mainframe? What is left to say about the IBM Mainframe? Intro L ad te trs hw ofsn! l of All confusing! how terms, other and EWLC duction hey try to provide the same service level as level service same the provide to try They g aalblt i a itisc component. intrinsic an is availability High T ake a look at the at look a ake Having of ourexperience. or pleasant results complexresultsits the implementation. Weyou deliver unexpected field, surprises highlight the gap between a brilliant idea and the in met cautionor comescriticism fromeach difficulties actual platform relevance. its validate Mainframe experienced experts. many between debated been has and experience users real-life during from gathered results the investigations. their and Mainframe Group CRIP Work the z/OS between meetings many from very experience. aspect; a important on insist Let’s solutions. practical tool a suggesting deliver to is Paper White this for objective Our see thatthere are manyofthem. we will review every single cost control method; you will wholeinfrastructure.the integratedin was it how and companies our in today Mainframe of position the In order to do so, we must first strike a balance, indicating just bearingitsfinancialburden. than rather provided, service the and cost its between the in it drive desired to way, monitor its but evolution and optimize machine the ratio the constrain to not is goal the reduction; mean necessarily not does control though, wrong it get not Do them. suffering of instead Software controlling keep investments; indeed, one must control to Mainframe costs prospects future even and leverages practices, best methods, are there that prove and right things set to intends Paper White This one application” era hascome to anend. rather than Mainframe itself; the glorious “one difficulties server, one OS, licensing software generate which virtualization and consolidation be to seems then issue A complexity.of lotmulti-core processorsa add job. full-time Systems are following a the same path; virtualization and almost Distributedregrettableagain then but indeed, situation is contracts hence, negotiating Software and misleading; managing rather analyzing, be understanding, can acronyms these very finding was actually implemented to implemented actually was finding Every he following commentary comes commentary following The c cmet ec nuance, each comment, Each his paper is the result the is paper This he true The Then

PAGE 5 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 6 1 2006 and2014. the average that shows share of Mainframeanswers only the complete applications would of decrease from is 28.6% which to average 19.25% platform, between overall z An the significant. on applications their of 60% run which of two 2014, and 2006 between constant remain would and stayed applications Mainframe of share answered) 2014. their by onlyapplicationsMainframe fully of all who stop to 11 plans them of of one only out furthermore, share; (6 their reducing are them of Most applications. only Mainframe of share their increasing are respondents the of none anyone; surprise not will This reducing very slowly. of potential onlyare applications hybrid whereasrarer, the becoming rapidlyare applications only by is it as threatened Distributed extinction, Systems? We facing observed two Mainframe complementary trends IBM about this; the Mainframe Is 1.2  to comparison in Mainframe of position relative the Distributed Systems, thesituationissomewhat different. at looks one if However, one, usually useitincreasingly. consumption. constantforecast firms year.two and Lastly,a 5% year,than less by but per consumption, their 15% decrease to and plan firms three only 5 hand, other the between On them. be of three for 5% will below increase the respondents, the of For increase. six consumption CPU a is trend general the future, foreseeable the In only one uses less than 50 MSUs (its Mainframe hosts about makes them half big users whereas of the majority consumes between 200 their and 500 MSUs; applications). which MSUs, 3000 and 1900 between consume surveyed people fifteen of out Four aredifferent.verysituations concerned, the is consumption averageMSU as far As will continue to bethecase for the2012-2014period. respondentsforthe used coreAll Mainframeis whose businessit applications said other computer resources by2014. complexwith Server z its integrate to plans respondent one and isolated, platform resources are integrated most of the time; only two respondents keep the Mainframe Furthermore,applications).computer businessmanagement other and Mainframe foronly z System used (who exceptingone surveyed,people the of all for true was The 2010 survey carried out by the Work Group of CRIP’s members showed that this Mainframeplatformthe Nowadays,for still used organizationsis an core business. 1.1 Role oftheMainframe intheeyes ofCRIP’s Members within aCompany “z” PositioningandStrategy (Windows, UNIX,Linux) The Position ofSystem zvs. Distributed Systems herefore, the Mainframe platform is in good health, those who have who Therefore,those health, good in platformMainframeis the Threerespondentstheir surveyed statedthat different sectors. This evolution is seemingly a consequence of the different business demands for the whereas usage, General Industry sectors are reducingincreased or stabilizing usage. for plans even and Mainframe the of use the question T the concerning opinions conflicting to theirrespective business sector. two accordingplatform, are Mainframe IBM the of future the and purpose the thereimportance, members, CRIP’s Among 1.3 System z the Mainframes consumed as“little” astwo Distributed Systems clusters. area, while running 25% of the applications. Similarly, regarding energy consumption, only 135 of the 4000 m² of the computer room, that is to say 3.5% of the whole surface occupied Mainframe the company, their Workin Group’sthe that statedmembersof topical are problems occupancy center environments. of One frugal most the is Mainframe the that mindful databe issues,let us and concerns energy when times In eachof on these environments indicates thattheMainframe usesvery running little storage. ofapplications percentage the with figures these comparing Mainframe, versus 72% to SAN and to15% to NAS for Distributed Systems. capacity storage their of 13% allocated users Mainframe Systems. Distributed for NAS to 21% and SAN to 75% versus Mainframe, allocatedto was storage of 4% from valid); be the still to respondents, seem they figures but (fromstudy, 2009 the Regarding storage for example, the CRIP’s Storage Work Group provided the following reached andthelow storage quantity itrequires, allneedto betaken into account. levelsecurity forexploitation,resourceprovided,high high the the leveloccupation spent time cost,parametersthe unit many as the to such addition Mainframe;in of a cost focused approach is sometimes too restrictive to prove the economic benefits Systems. Distributed than time-consuming less is administration its and people less needs the internal discussions of the Work Group reached the same conclusion; Mainframe figures.these of allowaccurateusage not respondents did of number low the environment; Mainframe the to allocatedmanpower the and budget IT the proportion the of obtaining about We question difficulty reliable the had answers to middleware servers ontheotherhand. to manages z System collaborate that with Distributed Systems and by becoming data continuing, servers on the one is hand and System Information whole the into the 2009-2011 and 2012-2014 periods. for 36% at stabilizednow is it 2008, and 2006 between 40% constant.was Where,it be to tends Systems, Distributed on partially and Mainframe on partially running Nevertheless, it is interesting to notice that the share of cross-platform applications, Systems increases. running applications of share the in drop solelySystemwhereason z, share the solelyapplications running of Distributedon a undoubtedly is there whole, the On o cut a long story short, the Banking-Finance-Insurance sector does not plan to plan not does sector Banking-Finance-Insurance the short, story long a cut o Banking/Finance/Insurance vs. General IndustrySectors he Mainframe still has a reputation for being expensive though. Indeed, though. expensive being reputationfor a has still Mainframe The This shows that the integration of Mainframe However,all Therefore,

PAGE 7 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 8 operational problem. possibleas much as from Mainframethe platform. by possible eradicate is orderto in particular) in environments(Java trendand tools deployingmodern this of impact the Reducing older. getting are teams an issue; is pyramid age the developers, Mainframe recruiting on go which plentiful. morecompanies and cheaper are skills Windows or Linux while skills, of lack for development Mainframe droprarer. companiesbecoming areSome skills These evolutions also generate skill issues. the Mainframe isoften shorter thantheactualmigration length”. As one of the members summed up; “the life length of a CIO who wants to eliminate outrageous. be would platform the leaveforced to being of cost the that underline to have sometimes teams Mainframe promise. application consultants non-Mainframe an what move to hard despite System, Distributed is a to MIPS of it thousands or hundreds consumes as which though, slow rather is phenomenon This is phenomenon this workloads; new without considered by a member of the Work Group as the “future death” of the but Mainframe. power, installed of increase As discussed earlier, this can be shown by some of CRIP’s members’ with a gradual the infrastructure, but a wish not to provide the Mainframe with additional functions. of renewalregular and applications and softwareexisting the of maintenance with wish to leave the Mainframe platform, since these strategic decisions run alongside (bespoke)applications.replacing homemade of aim Mainframe either rewritten applications the or replacementon software run products, to with the not is principle in which migration”, “soft towards be then can goal applications. existing of growth the for except applications, business core for homogenization, evenenvironment, Mainframe the within applications machines new of developmentexclude towards oriented often decisions, management’s IS generalthe IndustryIn sector, Distributed Systems are increasingly important. manipulates centralized data. run middleware layers (application servers) as long as, once again, such processing Mainframe is expanding of in these use sectors. Additionally, the the Mainframe thus is platform:also used to the on residing already longaccess, data as on depend Mainframe they as the for applications new of deployment the encourages fact platform. the in deep-rooted still are and evolving,keep ago, decades two or one written were which of some applications, COBOL business coreFurthermore, MainframeIBM the using platform centralizeda as data server bestthe is solution. highly is centralized, frequentlydata saved and maintained for perpetual integrity.that In this context, important is huge it Indeed, activity. core of the still manipulation is quantities date transactional sector, Banking-Finance-Insurance the In Everyone acknowledges that Mainframe Howeversolvenot does this the owever, we did not observe a observe not did However,we e in Even This The The n uut 09 IM rsne a e cmeca ofr eiae t tee new these to dedicated Solution offer workloads; commercial new a presented IBM 2009, August In end convince users. to enough necessarily not is this but generated, workload new this consideration. Java, XML and DB2 specialized processors lower the financial impact robustness.and T viability oftheplatform to continue. must adapt. of era the In limited. Web 2.0 and RIA (Rich Internet is Applications), these technologies have interfaceno future and user their and COBOL in written are programs Mainframe current of Most workloads”. “new call they what on insist IBM Firstly, relevant anddiversified enough to convince seniordecisionmakers? commercial offers, even too many and they sometimes confuse customers; are they Mainframe. the of longevity and attractiveness the boost to try problemand this facing are companies of lot a that aware seems IBM this renewal islikely to bedelicate are andstrategic choices must bemade. they and old getting are areapplications missioncritical,these naturallyrenewed.of mostbeing that Given platform this on hosted applications the However, businesses. big the of mostfor high, still is applications Mainframe of number The 1.4 Evolution Strategies for System z Mainframe cloud, acloud whichcan managehighheterogeneousness. a as such obsessions trendy to adapt to datacenter-in-a-box,or a as stand to able the in extensions. x86 on zBX Windows run also to ability the announced IBM 2011, April early zBX). In (with Linux/x86 and AIX/Power (z196), z/OS-PR/SM are which technologies heterogeneousplatform same the on host to able is machine z Enterprise new The to watch theoutcome ofthislatest evolution from IBM. high a represents zEnterprise. released,the was machines which of generation new platform, Mainframe the in its both proportionof secret(keptthough) investing on keeps IBM the usageofnew Java applications withoutincreasing WLC ontheMainframe? authorizing by game the change evolution this Will adjustment. Pricing) Workload (WLC). Charge License traditional hosting Software“qualifying” productsWebSphere)of (E.g. usage the existingpartitions on In 2011, a new billing offer focused on a premise of co-localization. and resource dedication theexception. rule, the is resourceutilization where platform, Mainframe the of functioning very projects (E.g. BI, SAP, CRM, et al.) with dedicated resources, which goes against the likecompetitivespecificallyreservesa and IBM offer,newto but usage targetedits Maintenance,whereasand previous the zNALC forSoftwareofferwas only. looksIt echnically, Mainframe is still one step ahead in terms of virtualization, scalability virtualization, of terms in ahead step one still is Mainframe echnically, This is why IBM wants Mainframe to run these new applications, for the However,comparedcost,its when a DistributedSystems, is x86 to e anrm tu bcms hprcnoiain platform, hyper-consolidation a becomes thus Mainframe The P monitors like IMS and CICS without increasing the Workloadincreasingthe without CICS and IMS likemonitors TP Edition. hese products are said to be eligible for IWP (Integrated IWP for eligible be to said are products These his financial package includes package financial This Hardwarea 2010, Softwaresales.In and here are many technical and technical many are There owever, we continue we However, rwr, Software Hardware, This offer allows

PAGE 9 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 10 original version”. of the use promises not would they why wondering up the end wanting will CIOs usage, by Mainframe because limit to and operational reversed and realistic be being from far may are Systems trend Distributed “this that replied members Group’sCRIP the of One engineering. computer of “mainframization” general a or rates. consolidation and virtualization high have which distributed) or (monolithic systems big about talk to Mainframe. IBM an like just run evolveand to platforms other preparedfor get requirement toa but usage, Mainframe to back going towards evolution trend a expect not must logical.more seemed datacenterSystem Distributed the in activities Linux consolidating datacenter.Finally, Systems Distributed the from for such usage, and that such operation would require a partial infrastructure move designed not was datacenter Mainframe the because value, added little very bring consolidate to Linux possibility workloads the in their Mainframe and datacenter; z their conclusion System was for that this Linux would of benefits the studied They is platform consolidation overall which has two datacenters, one for Distributed an Systems and one for the Mainframe. as Mainframe company a in WorkworksGroup the of member One facts. of by inhibited sometimes use the doubt, Without to befollowed bytheirorganizations. of matter a also administration team takes control of various systems; a path that is not really but likely validated), be must platformimplyzEnterprisewould adopting Indeed, morepolitics. lessor Mainframethe that the of capabilities real (the As the CRIP Work Group underlined, this evolution is not only a matter of technology he termsThe MainframeMainframe or areUNIX x86 more moreand used ence, eventually, there will be Mainframe-like systems, Mainframe-like be will there eventually, Hence, his case shows that one that shows case This 2 study completed in2010bythe15members oftheCRIPusingMainframe . a from principally come follows what in appearing data and figures the Attention: 2.1 Technical terms ofreference z Platform O rahn (ht) pc» o ovninl eea Proe rcsos and processors, Purpose General conventional to space» «giving investment, (white) modest a breathing for allows processors specialty of addition The implementation matter, specifically in France. whether this IFL is used in Production or customers have atleast oneMainframe IFL(late 2010).ButIBM doesnotspecify Globally, IBM reported significantly more flattering reports; approximately notable 30% of one (with customer no and nature exception) hastaken thestep ofdeploying style» Production onIFL’s. of «exercise an maintain they processors, (Linux) IFL for As (Java). processors zAAP used lesser the by zIIP processors (DB2 DRDA) are the most commonly used (~50 % users), followed information ontheirmodeoffunctioningandadvantages). lesstimes comparedwhen classicalwith processors formore3.2.2 chapter (see 5 about is cost purchase their Moreover, et cost. software MLC Linux, adding without Java, al) DB2, (E.g. workloads of types some support processors These Specialty Processors server charging mechanism,AWLC only appliesto MLC. forwardcarriedsoftwareratio O (MLC invoicingforIBM MSU + by IBM. It is also necessary to note that the older server which changed the MIPS/ made announcements optimistic the of spite in generations, previousserver for calculations made, the benefit for the client is of the sameform orderof a asnew thatsystem determined of invoicing called AWLC, replacing the classicThey noted that forVWLC. the z196 technology, Forthis technology dividend bonus takesany the • Dividend»)@~3to 5%(IBMindicates ~10%)from onegeneration to thenext. («T invoicing softwareMLC of time the at technology new for Bonus • • Evolution (age) of servers, 5 years since their introduction (General Availability). •  primary reasons: these for current, as classified be can technologicallyPlatform z Thereforethe replaced has 30% oftheinstall base. server z196 recent more the then since Although (z10). iteration At the time of the study, the installed z System servers are generally of the latest Infrastructure 2.1.1 Servers Concern for perpetualinvestment (I.E.for themost recent technology). verview T est. There is however a delay in this IFL T C) while the new whilethe C) echnology

PAGE 11 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 12 within some sites, used bysomeolderandslower devices, currently still inuse. ESCON connectivity is clearly losing ground in favor of FICON, but remains present Connectivity Protocols already in inherently satisfactory which demonstrates ahigh rate ofconsolidation. environment entirely Mainframe are a architectures in current performance, of that terms true also is It trends. will functions these more«fashionable»other environments worldas Mainframe storage,the in later appear of evolution the like that bet safe a like seems It flows (8Gbcurrently, 16Gbproposed) are onthe roadmap. no major technological innovation is delivered by these IHV’s. Only increased data platforms(FCodistributedother worldwide), E on contrastobserved In activity to T Infrastructure 2.1.2 SAN&Networks its limitations ifwe decideto adoptit. brings. it simplification technical usage. heavy for the represents that MLC budget the of volume overall the given overlooked, be not MLC via simple consolidation. in decline ~2% a seen architecturehave this implemented have that Customers any preconditions, thesametariff benefitsasthepriceplex architecture. this solution can certainly not be ruled out, especially since it can deliver without management. outage including the increased reliability of equipment. given benefits, many have now can ago, yearsfew a onlyunthinkable (CBUs). was which processor backup server-only a with site backup a and site single-server a witnessed; sometimes architectureis this to alternative An software billbyabout30%compared to classical/conventional architectures. metrics (aggregated PSLC). This is to derive benefit from tariff conditions applying to IBM MLC billing software The overwhelming majority of users adopted Sysplex architecture in a «priceplex». Sysplex Architecture DB2 andJava. disappear in favor of zIIP. Finally, there is a convergence of types of specialized processors; zAAP processors changes are notalways easyto anticipate. loadspeak of 4 (Rolling peak load may be generated, as high as secondthat which a has Similarly, been billing. moved. generating load peak the on impact no or little have not always follow, because the offset of activity in specialized search engines may correspondingthe MLChand decreasethe softwareother in invoicethe does On is agoodresult, ifthenature oftheproduction workload iseligible. therefore allowing to differ growth investment. ’) hr te akt Boae n Cisco. and Brocade market; the share (IHV’s) vendors equipment SAN wo our) resultsHour) from complexa fragileand balance, where i acietr i itrsig n h eooi and economic the in interesting is architecture This The latter can now accommodate either workload, both This technique makes it possible to reduce the monthly This perceived low percentage should nevertheless ven if it is not currently the most widespread, most the currently not is it if Even owever, we should nevertheless be aware of aware be nevertheless should we However, However, it can cause problems of usability Therefore from this viewpoint, this i solution, This The evolution In terms oftechnological developments, thetrends generally observed are: Technological Developments market; storage followed byIBMandEMC². Mainframe the share players Three Infrastructure 2.1.3 Storage controller communications 3745 the types have totally disappeared, infavor ofT and SNA protocols, network of terms In uncommon amongusers, for thefollowing reasons: IHV’s.storage technicallyall possiblefor is Pooling Pooling withotherenvironments range ofsoftware usedisimportant. be can and negotiable easily and variety the when suppliers, differences price between mostsignificant sourcesof the are costs software these Conversely, storage, of both interms ofinvestment andoperating costs. cost the increase greatly and significant, is functions these of cost now staples of Mainframe storage. On the other hand, according to the IHV’s, the products is both widespread and somewhat generic. PAV/ HyperPAV functions are functionality. asynchronous) or (PAV,synchronous (copy performance security to dedicated mainly is software This Embedded Storage Array Software (Microcode) •  •  (3 price its but appearance, its made • T(SSD) storage high-performance he •  •  g-n soae usage). storage (High-end storage Mainframe expensive most the and Systems Distributed between GB per gap price significant and Mainframe storage arrays, excepting connectivity, there is still a commonplace Conversely, while there are many similarities between Distributed Systems and •  •  •  •  disk IHV’sdisk agreelow-endSA that all but world, Mainframe the in commonplace evolved,is therefore tiering Storage have disks midrange requirements. GB+) access/performancelow (600 for especially storage, overall of cost the reducing capacity larger Conversely, very specific/nicheapplications. for low, as usage its confines now storage) conventional than higher times Replacement ofFiber ChanneldiskswithSASattachment. are cheaperto produce, for similarly equivalent performance. which disks, rpm 10,000 inch 2.5 with rpm 15,000 inch 3.5 of Replacement storage remains essentially ashigh-endstorage. 1-2). Level to migration DFHSM for perhaps (except rates, I/O higher requiring usage, Mainframe 0) (Level storage primary for Organizational (technical teams) issues. Indeterminable cost benefits. Complexity ofadministration andassociated risk. Storage. Distributed Systems storage volumetry is not on the same scale as Mainframe T A type solutions aresolutions generallysuitabletype not A CP/IP andOSAcards. owever, usage is generally is usage However, S h cer leader, clear the HDS Therefore, MainframeTherefore, prA) n data and HyperPAV) he usage of these of usage The i i otn the often is This

PAGE 13 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 14 years, delivering geometrieswithincreased capabilities: The geometry and granularity of disk storage has changed dramatically in recent Mainframe diskstorage geometryevolution vr ie h 39- dvc cpct hs eoe o lw o et usage meet to low too become has capacity device requirements, generating thefollowing consequences: 3390-3 the time Over volume isstill very rare for thefollowing reasons: allowsforextreme/exceptionalthat geometry new singlelogicalThis a as usage EAV ( Extended Address Volume) –SpecialCase: T oday we can identifysometrends: g System E.g. ( usage specific for reserved typically are devices • T 3390-9 he • The legacy 3390-3 devices have almost disappeared in favor of the 3390-9 and reached thisprice harmonization. important customer cost reduction focus, for customers who have not already providers. with gap negotiations during price worlds two the the between reduce to Directions Puchase of inability the of consequence •  •  •  •  • The physical storage reorganization must be accompanied by a reorganization •  • The increase in multiple-volume and architectural multiple-extents (E.g. files and pool the associated storage the for required volumes of • T number he •  storage pools. problemssignificant utilization canpose which smaller disk the users of for is likely that their use will become widespread, in spite of their unit capacity, controllers. it Nevertheless, host generation latest the from benefitting not when especially device, this with encountered be can issues performance disks). commentary indicates limited andoccasional usageofthe3390-18device. Anecdotal pools. storage the of most occupying devices, 3390-27 especially Limited experience and feedback from thefield. Migration istypically cumbersome andnottransparent. Using EAV induces manysoftware pre-requisites. will beniche,andonly usedfor aminimal numberofapplications. logic mentionedpreviously. management policy DFSMS the redesign to necessary more even it makes which geometries, disk two of maximum a manage desirableto is Finally,it the including the Group) thatare now ofgreat policy), assistance. to due DFSMS Additionally, pools. (E.g. decrease of granularity, storageDFSMS features have evolved associated management (E.g. Overflow Storage the data of reorganization logical the of IBM, (E.g. sellers the from tools Innovation, TDMF,etal). movement data non-disruptive using disk renewing controllers. when Nevertheless, it ideal is possible considered to achieve be this storage canreorganization reorganization storage Disk increase inoperational incidentsrelated to lack ofspace (E.g.x37abends). constraints). geometry, especially the 3390-54 device. device disk “traditional” to suited not are applications of number limited A he more recent 3390-54 device is still not widely used. Some minor Some used. widely not still morerecentdeviceis The 3390-54 Thus it is anticipated that This remains an remains This EAV usage replacement intentions are perceivable: renewal/ and ageing, are equipment associated and functions Oracle However, Oracle (formerly Sun/StorageT Infrastructure 2.1.4 Backups (site). copies without the backup need to physically transport generating physical tapes technological to another distant,location replication.and local by both asynchronous sites, separate possiblefor and allow techniques made synchronous remote was in systems developments these for breakthrough The to managefrom theperspective oftheuser. Generally, the newer systems regardless of the underlying technology are simpler advantages are: capacity. variable of drives tape Mainframe emulating controllers technology made its appearance in the Mainframe world. Mainframe the tape in appearance virtual its made addition, technology In exist. to continues and years many for commonplace T Technological Developments p vrulzto ascae wt lrecpct ds cce a been has cache disk large-capacity with associated virtualization ape • EMC •  •  • Reductionoffloorspace andenergy consumption. • Tpossibilityhe deduplication.of •  • Eradication ofDFHSMML1. • Elimination of the problems associated with filling physical tape media, which • Cost. •  positioning to thatofIBM. of terms in similar somewhat solution, interesting and noteworthy a offers leadership in technical and commercial developments. and direction of lack a for criticized often areOracle relationship, business the in confidence of lack this to addition In sectors. market other for policy trade controversial a to related perception brand poor from suffers Oracle + Disk (full conservative virtualization). and stable both was usage IBM Generally will beinteresting to monitor thefirst field experiences. It suppliers. from contrary the to assurances despite storage, conventional comparedwith problematic,morewhen be reconstruction candata data, of lossphysical of event the in that sight lose not must we Additionally, made. any In be can managed. decision a beforeessential datais types data the of analysis detailed a of case, nature the to due environments, other with comparedwhen Mainframe the interestingforless much is technology this Improved backuptimes. continue to grow in capacity. LUMINEX, deduplication) inthe2011-2012timeframe. (E.g. announcement an make to expected is but document, market, backup Mainframe this writing of time the At virtualization solutions. below). (see deduplication as such technologies This substantially simplifies and improves the reliability of DR/BC (Disaster 2 recently entered the Mainframe backup market, focusing on innovative ek) is the major player in France, followed by IBM. Therefore virtualization is somewhat mandatory. Evenmostthe ardent supporters that admit S s o crety cie n the in active currently not is HDS V only provides disk providesonly IHV This his is actually SA actually is This However, Oracle still e primary The These T ape T A

PAGE 15 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 16 g ~0 k fo primary from km ~500 (E.g. sites backup remote have also customers Some a highlevel ofsecurityandcontinuity ofservice. center data the use of synchronous data replication two systems between two sites, safeguarding of consists architecturecampuses (two local sites separated encounteredby less than 20 km). frequently most The Technical Architecture 2.1.5 Technical Architecture &Organization on disk(eitherintermediate orpermanently); butfor how long? storedactually is data wherethe types, required, drive legacytapeis supporting drivestapefor compatibilityreasons. traditional emulate Systems, Distributed as considered be can that controllers interesting offers technology tape disk systems, these costs.On lowerbackup to users allowinginnovations,many virtual that found study our conclusion, In (E.g. CA-MIM). management pool tape for products use to requirement the eradicating drives, Additionally, all virtual tape systems offer a sufficient number (I.E. 256+) of logical even essential when: increase. to continues tape physical of capacity the Similarly, systems. tape virtual tapelessto performanceequivalent evolve, to continue A and caches virtual media capacity high with combined and systems tape physical conventional However, most commonly usedarchitectures: the describing diagram a is Below procedures. Continuity) Recovery/Business • Data capacities are significant (E.g.nTB+). • The removable (physical tape) media is essential for job processing or security reasons. Figure 1:Sample HighAvailability Mainframe Configuration his is the onlythe reasonis This emulation tape that hese systems remain competitive and competitive remain systems These .g. IBM, Oracle) provideOracle) TL’s IBM, (E.g. This architecture allows made inthisarea, primarily for storage andbackup,including: T primary site. the of destruction total or partial the from resulting DRP) - Planning Recovery normal operation, and used in the context of a Disaster Recovery asynchronously. performed is site backup the and primary the between data of Replication site). echnological and economic considerations influence the architectural choices architectural the influence considerations economic and echnological the synchronous copy (Figure 3). and storage primary for both copies, disk full dedicate to is solution best the recognizedwidelythat now is it Consequently, (IBM). GDPS as such solutions automated with incompatible is it Moreover, unavailability. storage primary lose to tends favor technique because it overly this complicates bottleneck,the emergency procedures a associated longer with no is bandwidth that Now •  called cross synchronous copy (Figure sites. 2). two the of other the to distributed also bandwidth between the sites. In this case, the synchronous data copies were divide the data between the two sites to minimize the data to customarygenerally flow is architecture,siteit andprimary two local associated a of part As hese backup sites are actually fallback sites, inactive during inactive sites, fallback actually are sites backup These Figure 2:Cross Synchronous Copy–Classical LegacyArchitecture Figure 3:Campus ModeSynchronous Copy–New Architecture his configuration might be might configuration This Event (Disaster

PAGE 17 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 18 latter. the chose study this in participating companies the half over just total, In help you). will heaven and yourself (Help solution outsourcing an for arguments all remove Compass) (E.g. «Benchmarks» Performance Standard good Moreover, management, direct of favor in regardless ofotherconsiderations. decision a dictates and essential, as some by control of their information system. The underlying position for firms considering this proposition regards the overall Outsourcing: Y cost savingsdelivered bysharingcomputer center runningcosts. of becausemakers, decision many of attention the holds problems,but security businesses pooling their resources with other data centers. commonplace. and used widely are subsidiaries their and companies (hosting) computeroutsourcingFrance, the In Datacenters Disaster in degradation Recovery performance mode.Therefore thismodeofoperation isgenerally more expensive. slight tolerate to or installed, powerful Disadvantage: this mode requires to bring all storage arrays in line with the most •  • The asynchronous duplication of backups is becoming widespread, gradually -  e anrm piaiy ss ihed ik arrays. disk high-end uses primarily Mainframe- T he -  •  and widespread, becoming is deployment copy synchronous of • T type he to addanextra layer ofabstraction. a unnecessaryis completelyit and thus nativestandard function; a (LPARs) is with partitions multiple from and array Mainframes more or one from differentdata to Access for levels performance virtualization layer. prioritize to necessary anrm dt vlms r cnieal lwr hn n h Distributed the in than Systems world, andtherefore lower thenumberofdiskcontrollers also. considerably are volumes data Mainframe way to anticipate radical changeofthelandscape intheshortterm. no is there backup, and storage for but somewhat, proposition this modify other environments. with connectivity of terms in especially compatible, always not is hardware result.Moreover,would constraints Mainframethat and risks of view lowin are similar. uncommon, although the basic components, including storage requirements quite is platforms non-Mainframe other with equipment sharing Generally, these while links, telecom replications and are mostly for remote sites. equipment backup for costs lower the by replacing the traditional physical tape backups. ite neet o te B Mifae ltom piaiy o te following the reasons: for primarily platform, Mainframe IBM the for interest little has DistributedSystemsworld implementedthe as in Storagevirtualization cost, recognizing thereduction ofdiskstorage costs. incrementalacceptable and lowrelatively a proceduresfor emergency and data. all for used be to tends es orNo The savings achievable by this type of pooling are generally quite The emergence of hybrid zEntreprise (zBX) servers may his significantly simplifies the management the simplifies significantly This The risk of power loss is an aspect considered here are also a significant proportion of proportion significant a also are There This has been made possible This obviously poses eeoe t s not is it Therefore availability are reduced, primarily dueto: Please refer to section 2.1.1). (priceplex, reasons financial for users all almost by deployed although solution, requiressolution Sysplexarchitecture.availability continuous The Continuous Availability &HighAvailability 2.1.6 Business Continuity year software contracts. multi- like just budget, multi-year predictive of certainterms in a logic same to the followsextent It customers. some for alternative attractive an becomes infrastructure Mainframe of rental the it, from far majority, the not is it if Even Infrastructure: LeasedorOwned most popularimplementations, delivering a«highavailability» service, dueto: Logistics/ Banking, (E.g. usages application T of minority a to limited typically but sectors, business all in found is availability continuous requirementfor The down time)Mainframe isperforming very honorably inthisarea. platforms. all for exists architectures Nevertheless, this issue of additional cost and complexity associated with cluster online reservations (E.g.Air/RailTicket) business. is found where thespecificbusiness needsor conditions dictate, for example, an deployed when a compelling need is evident. In is particular, which this solution, type this of of implementationarchitecture the for obstaclesare factors these All emergency procedures are notalways properly managed. Experience shows that in the event of a major incident, the associated restart and that face major risks. and DRP, a have systems information their surveyedof continuity business the accounttherefore companies into take the all that reported study our Firstly, DRP (Disaster Recovery Plan) maintenance orminorfailure outages (~5orless peryear). advancesthese of All allow for considerablea reduction overallthe in of number lcmuiain/tlte, t l) Cascl rhtcue sil ean the remain still architectures Classical al.). et elecommunications/Utilities, • The overall additional cost of such a high-availability solution, • Thecomplexitywhen ofreal-lifecompared (Business AsUsual)management. • Thecost ofadditionalresources required for thisarchitecture (~10%). • Thecomplexity ofimplementation andoverall cost. •   • (I.E. withoutservice interruption). dynamically implemented be can changes technical of number increased An in terms of intrinsic hardware reliability (MTBF – Mean Reduced hardware failures linked to the progress made during the last 10 years with aconventional solution. However, the requirements for achieving continuous owever, the zSeries (z represents zero represents (z zSeries the However, Time Between Failure). owever, this However,

PAGE 19 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 20 2009 (including $3 Billion for the for Billion $3 (including 2009 in Billion $8.5 totaled systems operating Mainframe and markethardware the for For reference we are providing some context information. According to IDC, revenues Background: required, infuture to-do andcomprehensive studies. technologies (IMSsoftware). We must capture thisinformation withtheduediligence to detail some information, particularly for small ISV’s and theactual use of certain Due to the lack of experience and time, it appears that our investigation has failed DBMS, Transaction Monitors, Monitoring Tools, Etc.) andnot the application software. ofin the CRIP onlysurvey Spring 2010. The survey covered infrastructure software (OS, The following information summarizes the responses from the 15 companies participating W 2.2 Mainframe software vendors (ISV’s) analysis &positioning arning: •  for performance Recovery Disaster the harmonize to need a • T remains here •  Nevertheless, we observed thefollowing: minute. RT an and sites) mode campuses two (with RPO time zero a with availability high deliverarchitectures Some others). for model a as serves CurrentlyMainframethe platform probablyis bestareathe this of in one even(it reliability, in advantages demonstrate they (Please performance (RPO/RT particular, the In discussed 2.1.6). from previously Section greatly architecturesee benefited and obviously developments have technological Plans Recovery Disaster These been typical for thissector to extend suchprocesses for ITOperations teams. therefore has It tomeet counter-example. notable a providesand mandatory requirements, regulatory is paramount training personnel flight of Plans. importance inthe Recovery Disaster improvement for a margin of significant performance a perhaps concerns. still rank is second there only However, are considered, always generally factors, These - T - Theperformance impactwhenoperating instressful conditions. - Staff training. ial, hl ams al opne ognz Dsse Rcvr tss to tests Recovery Recovery test performance alsoconsiders humanfactors suchas: Disaster organize companies periodically check the reliability of procedures, all it should be noted that Disaster almost while Finally, 5-10 for restarteddays? be cannot components application non-Mainframe other 24 within Mainframe the datarestarting in a there is of interest what loss center, the following platforms, multiple across distributed are typically applications when juncture a At platforms. heterogeneous and different via SLAmetricsare relaxed, from auserexpectation viewpoint. Unavailability outages discovered during Disaster Recovery testing, documented eamwork, etal. O) andsimplicity. uropean geographical area), whereas software whereas area), geographical European e iln idsr, hr the where industry, airline The O of less than one than less of O ours if Hours small, creatingsmall, fewvery software productsexpandedportfolios,by sometimes with and large both ISV’sdisappear, of number large a seen has consolidation ISV This stabilizing for several reasons: before1990’s the since dramatically changed has market software this Moreover, not finding thatIBM if revenue share wasslightly higher. observation, this confirm least at to seems investigation our while market, estimated, again in 2009, that IBM controlled about 40% of this Mainframe software IBM is very discreet about breakdown of their revenues by product line. server (hardware) thusequates to software witharatio of~1:3. a Purchasing year. same the in Billion $24.5 for accounted • The buying patterns for major companies have evolved along with the introduction the by exacerbated was ISV’s small with associated risk of • T perception he •  in or infrastructure, their streamlining consolidation, company • Tof impact he Systems Distributed • T in he dramatic increase of Mergers increase & Acquisitions (M&A) in the software industry the context; technological general • The shelfware products create extra maintenance revenue. these that hope the in products,“top-up” deliverISV’s to major the encourage customer. the by used adoption not commonplace software “shelfware”, a of have now we strategies, purchasing new these of continuity on stability, of solutions, again focused to the rather detriment unless but of smaller considerations, response, ISV’s. technical As generic specific a consequence by more dictated a to isolated purchases, based time logical upon concentrating now are companies Our benefit. maximum the delivering for portfolios, software enough large with ISV’s to applied only but leverage, of multi-year global negotiations. for product compatibility. adjustments, necessary and minimum the providingonly by or products, their burden on some software vendors, even if only to retrofit these innovations into every six months. evolvewould which OS/390, including products,software Mainframe some for IBM. versions of turnoverrapid a perform to by 1990’s,decided late IBM the in requiredIndeed, platform Mainframe the for evolution of pace accelerating provide support,innovation ability,etal. to ability durability,ISV’s; small with inherent risks alleged the of do because so to reluctant generating been have customers Mainframe large suppliers, while supplier, small a with onmajor engage to dares size modest of customer Mainframe A ISV’s. small to but appeal their limiting relying fewersuppliers), few or (single become concentration of important, phenomenon organizations customer more consolidation, significantly of result a As account (customers). centersenterprisereducedusers,and which overallthe Mainframeof number data of number the reducing (M&A), Acquisitions & Mergers with connection portfolios. software their expanding while (ISV’s), suppliers of number the reduced has computing hasdecreased thedemandfor Mainframe solutions. This state of affairs that lasted a few years was to inflict a heavy They were aimed to provide greater bargaining u goa negotiations global Thus rpa Mainframe European However, IDC

PAGE 21 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 22 SAS, etal). IBM, Compuware, CA, BMC, ASG, (E.g. vendorslarge very few a aroundstructured functionality. duplicate established: well typically are and areas functional small focused several or one have They The nicheISV’s: ISV’s: main 3 the as broad as perhaps areas, functional three or two with specialized rather but size, significant and presence global a with ISV’s several are Below The large specialist ISV’s: 2.2.2 TheSecondary Suppliers (ISV’s) •  •  •  •  The 3primaryISV’s: portfolio andfinally, nicheplayers. range and appear almost inevitable, those who have a more specialized software We can distinguish three categories of providers; the three who have an extensive 2.2.1 TheMainSuppliers (ISV’s) BMC has an especially strong foundation in the and Systems security Management discipline. of fields databasethe tools. However, securityremains acompetitive market. in position strong a has CA history, its of Because (I.E. CICS&TPF)anddatabases (I.E.DB2&IMS). represents only marginal activity, but also with transaction processingwhich monitorsLinux, except Systems, Operating Mainframe with supreme reigns IBM position ofstrength. category. this market, over intime they have established comprehensive offerings found and occupy a be can BMC and CA IBM, •  •  •  IBM offerings (I.E.CICS,DB2,IMS). environment), making Adabas one of the few competitors with the equivalent development associated (with monitor transaction associated and (Adabas) a market they Primarily France. in presence good a established in its home market, 4th in 4th market, home its in Software AG is a significant German ISV that is well established globally (2nd development application assistance. and Cortex of acquisition the with management with monitoring areas; three in catalog portfolio software significant a has ASG survey, our in represented poorly Although competitors. of devoid almost position a establishingnearly of point the to position, their strengthen and preserveto managed Compuware ago, yearsseveral areas, manipulation, analysis of incidents, et al. After aid, gaining the developmentleadership in of these tools debugging, trace the analysis, performance analysis, creation in of test cases, data presence strong very a has Compuware T dy h IV adcp i lrey tblzd ad tightly and stabilized, largely is landscape ISV the oday urope and in the in and Europe T op 25 globally), but has never has globally),but 25 op soial peet n the in present Historically O, rdcin (job) Production TMON, In order ofpopularity, IBMwithCICS, IMSandWebSphere. > For Transaction Processing (TP)Monitors > For Operating Systems importance: orderof in members CRIP the by used as softwarecategories, major Presentedby 2.3 Survey Results 3. Novell-Suse withtheirversion of zLinux. 2. RedHatwiththeirversion ofzLinux. 1.  W •  those thatwe have forgotten. our questionnaire. An upcoming future survey will review these ISV’s, highlighting of their existence and use of products. •  •  infrastructure, primarily disk from disk primarily infrastructure, residing on the Mainframe, such software is required for their associated storage and EMC² from storage with associated components infrastructure software technical of note make also we Study, CRIP this of subject the been not have they Although •  •  •  ial, e oe ta tee s pplto o sal pcaie IVs that ISV’s, specialized associated recurring maintenance revenues. small and base install of their on live population now and solutions, a common quite marketed is once there that noted we Finally, StorageT that owned Interpel (Pelican) and bought CFT (Synchrony Axway reign supreme for Mainframe file transfers. Indeed they are the company with tools for administration and handling. especially,DB2 and IMS marketdatabases,dedicated the of ISV toInfotel an is development aid,debugging,trace analysis, data manipulation,etal. Macro4 has developed competition for Compuware development tools, including perceived astheonly product onthemarket. manipulation, competing withtheIBMDFSORT offering. data and sorting of field the in player ISV party 3rd active only the is Syncsort Series Mainframe on z forplatforms. processing OS Linux their of versions specific of belong to the family of Mainframe specialists, but provide IBM with the blessing reallynot do that distributions Linux ISV’sof two areSuse-Novell and RedHat traditionally successful inGermany,SwitzerlandandtheNordic countries. management. It is surprising not to see this ISV in our survey, even if it is rather Beta Systems is a leading ISV in the areas of production automation and output arning Reprise: arning Systems. IBM withitsubiquitous andmultiple generations ofMainframe Operating ek). DS, plus backup software from Oracle (Sun/StorageT Oracle from software backup plus HDS, Our investigation failed to highlight small ISV’s which we know MC² and EMC² This was how we worked and the shaping of This ISV is also very Franco-French. DS and tape from Oracle (Sun/ Oracle from tape and HDS T ransfer), sometimes ek). Naturally ek).

PAGE 23 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 24 and CA securityofferings are now technically very similar. IBM Both fromCA. sharesback markettaken and up, caught since has IBM users. has taken market shares from IBM at a time when RACF was not satisfactory for its CA States. United the in successfulmore is ACF2 then France,but in popular very In this area, both providers IBM and CA are notoriously competitive. notoriously are CA and IBM providers both area, this In >  CA Scheduler, ASG-Cortex family, andeven lesser-known products, butstill used. other products we haven’t referenced in this document, but quite popular, including available. widely some ISV’s, credible from products good Thererealis competitionhealthy and market the in batchof schedulers, many with > For thebatch schedulers cases where the hierarchical database functions of IMS are particularly well suited. historicalan legacy, fortoavoidmigrationcosts,specific wish companies in that or We were surprised not to include IMS/DB in this list. years. the over competition eliminated slowly has IBM challenge, technical this passing By robustness. and performance features, of terms in advantages with technique differential a maintained has IBM expensive, although and challenges, early after develop DB2 on the z Series platform, IBM but SQL. also IBM with no all then partnered with Oraclewas other to complementthere with the major DB2 tool. By investing and heavily Operating to database Systems, relational a delivered yet not had IBM 1980’s,late the products.In other of fewusers a and (DLI) IMS/DB residuesof with Mainframe, including Adabas, Datacom, IDMS, and others. Now, it is almost all DB2, marginalcompetition. IBM has completely conquered this functional area to the point of only observing very > For RelationalDataBase ManagementSubsystems (RDBMS) accelerating the pace of its updates to products and investing in its own environment. by aside, competitionliterally the has swept IBM addition, In developmaintain. and to provedcomplicated and reliability of level high requiredproductsa the because A somewhat surprising result; originally there were some competitors in this market 3. CA withCA-7. 2. BMCwithControl-M. 1. IBMwithTWS(formerly OPC). 2. Software AG withAdabas(observed withmeritinthisstudy). 1. IBMwithDB2. 2. CA withACF2 andT 1. IBMwithRACF (SecurityServer). For Security Subsystems owever, 20 years ago there were several for the fortherewere databasesseveral ago years 20 However, op Secret. This product is still used, but as here are also several also are There T op Secret is Secret op > For DB2Utilities Systems andBusiness Service Management(BSM)solutions. Additionally,lastfewthe during focused yearshas Distributed ASG on appears that it from its status as a small independent ISV, during the times of whiletheyenjoyed circulation wide a increasing only fewa probablyyearshas ASG ago. suffered consolidation. fewASG so see to surprise a of somewhat is products.It where BMC are strong, neck and neck with IBM, now areas marketingfunctional threearethere general,the in Omegamon Monitoring Systems Candlez/OS regardto With > For DB2Systems Monitoring > For IMSSystems Monitoring > For CICSSystems Monitoring their core market. competitiveproducts, performanceof activity this but monitoring techniques, not is in particular,theOmegamonsoftware suite, returning to thismarket. CA hassome sophisticatedsoftware. Finally, directionits changed IBM acquiredand Candle, and for IBM to provide basic tools, allowing users to select third-party vendors for more enough was area,whereasit this in invest not did IBM years Formany Babbage. & with Boole leader from line product the the purchased previously remains having offering, BMC MainView their where sector, market competitive fairly Another > For z/OS(MVS)Systems Monitoring 4. ASGwith TMONfor DB2(observed withmerit inthisstudy). 3.  2.  1.  4. ASGwithTMONfor DB2(observed withmeritinthisstudy). 3. CA withInsight(observed withmeritinthisstudy). 2. IBMwithOmegamonfor DB2. 1. BMCwithMainView for DB2. 2. BMCwithMainView for IMS. 1. IBMwithOmegamonfor IMS. 3. ASGwithTMONfor CICS(observed withmeritinthisstudy). 2. IBMwithOmegamonfor CICS. 1. BMCwithMainView for CICS. 3. CA withSysview (observed inthisstudy). 2. IBMwithOmegamonfor z/OS andRMF. 1. BMCwithMainView for z/OSandCMF. Infotel MASTER-UTILproducts. RC/ IBM withDB2utilities/tools &Optim (observed with meritinthisstudy). (Platinum, Solutions Management Database DB2 Migrator, RC/Query,RC/Update, etal.) of family a with CA MON productsreferenced,TMON

PAGE 25 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 26 3. IBMwith SCLM(observed withmeritinthisstudy). 2. ASGwith CCC/Life Cycle Manager. 1. CA withEndevor. > For Configuration Management the existence ofoneAdabas userinoursurvey. list, this in appear relativetheir immaturity.of because perhaps Software AG’s presence not explainedby is does which assistance, development software for portfolio comprehensive a introducedrecently IBM note; sector.Please the into competition became a popular and niche supplier, where their positioning was successful, injecting at a time when many customers were in serious trading (financial) difficulties. aid. Macro4 entered the market with an offer that directly rivaled Compuware and this platform. We then find Compuware, with a full suite of specialist tools forMainframe development their forcompilers and languages programmingprovides IBM Obviously > For Application Development other are there although survey, our from solutions, suchasthosefrom CA, whichdidnotappearinourstudy results. products major two listed have We > For Automation for several tools decades, whichexplains thewidespread usageofNetView. provided consistently has IBM where area an is monitoring Networking > For Networking Systems Monitoring from oursurvey results. equivalent to that of CA, but we were surprised not to see functionally their toolsproducts of representedsuite comprehensive a offers FrancoFrench ISV.BMC small a is Infotel tools.developed internally its of youth the of representedpoorlybecause remains but recently,more market its this since entered has position IBM dominant Platinum. a of has acquisition CA which for market, active very a still is This 4. Software AG withNatural (observed withmeritinthisstudy). 3. Macro4 withT 2.  1.  2. BMCwithMainView AutoOPERA 1.  3.  2. BMCwithMainView for IP. 1. IBMwithNetView andOmegamonfor Mainframe Networks (IP). IBM (inorder ofpopularity)withCOBOL,Assembler, C,Java, FORT RAN, PL/I&QMF. ServicePilot withIPWatch (NBA,360)for z/OS(observed withmeritinthisstudy). (observed withmeritinthisstudy). Compuware withXpediter andFile-AID (DB2,CICS)family (plus GDPSfor Disaster Recovery). IBM withTivoli System Automation (SA)for z/OS raceMaster (observed withmeritinthisstudy). T OR for z/OS.

This ISV ISV’s andtheirrespective software portfolios. these remarkableforstability the of evidence again, Once bases. respectiveinstall products remains low, especially as BMC, CA and Compuware have maintained IBM their these of penetration the that evident seems it later, years functionality.Some introduced, were tools viable. Administration Monitoring Systemmanufacturer with the providingendowedCandle of acquisition the while financially therefore Database and IBM; and sound Development by coveredApplication technically not solutions previously with areas customers the in software of developmentinternal or acquisitions throughofferings, own their developto plans way so their tools were far too expensive. dishonest a ISVsbilledin faultof who the fault,but their not costprohibitive,was it being of perception a state of fromsuffered Mainframe the if seeming that stated this IBM equilibrium. threatening occurred, event significant a ago, years few A their qualityandrelevance hasbeenwidely proven. because persist and years, 25 or 20 15, deployedfor be toproducts for uncommon alone.technicalfunction on based ISV, unlikelyan displaceis to dictates that ISV’s are unlikely to change their business practices and so a customer with infrequent changes. Additionally, the maturity of Mainframe software products Mainframe is a mature market, with products proficiently deployed, honed practices,innovation. worthwhile little with slowly, very move Security and Management continuous technological innovations made to DB2, functions such as Configuration the to due active remains Administration Database of function the Whilesoftware. Firstly, please note that not all sectors benefit from dynamicSummary: tability innovation for Mainframe time. primarily via acquisition. market, the in dominance of position a gained has Axway previouslyexplained, As > For File Transfer SCLM offering remains particularly low. IBM the few of usage the and area, functional of this in products one no has BMC with management. history long a with Production/Configuration in strengthrelative with leader ISV an is ASG the products.existing is CA place, market this In 2. DDMSwithT 1. AxwaywithCFT(Synchrony T om The T om offering from DDMS has almost disappeared over ransfer) andInterpel (Pelican). T o encourage competition, IBM announced Therefore,not is it Therefore The

PAGE 27 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER 3

PAGE 28 z Platform CostControl jigsaw pieces are inplace. technical environment identified itsecosystem andwe of suppliers;allthe tions. We identifiedhave the rolewe outlined have the of the Mainframe, pictureThe Mainframe intheprevious hasbeendrawn document sec- two 3.1 Controling Software Cost help you explore new pathsorto consolidate your existing efforts. initiativebusiness.their in and imagination topics. their uses everyone other that doubt no and have We topical. and practicesextensive other certainly are there course, Of important (ifnotmajor)for mastering Mainframe cost control. as identified topics with results, associated their and implemented practices from member, group each of experiences the from conducted work group, Mainframe This identification and explanation for these themes are the result of work of the CRIP •Cost control for operations. • • This following chapter isdividedinto three majorsections,withseveral themes: models). Moreover, this approach requires companies to acquire and maintain and acquire to companies requires approach this Moreover, models). licensing their tocomplexity add to tends virtualization, namely world, Systems Distributed the in developmentsrecenttechnological the careful, be but ; world Systems Distributed the from coming licenses blanket (including models tariff simple and simplistic of favor business in Mainframe senior IBM the leadwith makers’part decision could which image, bad a platform Mainframe the gives customers. their revenuefrommost the gain to be could operandi contracts, and which models easily licensing confuses of their plethora customers a (See created Figure has 4 IBM below). years Potentially the this Over modus Services andSoftware Options), whichwe willdiscuss. undoubtedly is prominent most the contracts, software these contracts.Among negotiate not unfortunately,do softwarebut all havelicenses, companies user All products. software more or several to applying potentially margins, tariff trading greater offering agreements are contracts license, a in Please remember that each software product is subject to a billing model described available from IBM.Aproactive approach isneededto choosethebest options. contracts associated and licenses software of range the know to important is It 3.1.1 IBMContractual Arrangements Cost control for infrastructure. Cost control for software. However,wematters the that hope discussed below will SSO (Enterprise ESSO i sbet is subject This his practice This en nasltd ihn goa bsns contract. business global a within encapsulated seen often is competition metric billing of type this fact in tobecause important,«theoretically» are open theoretically O an per functions as charged are for software, Mainframe Other the cost borneby theMainframe reduce platform. visibly to metric, billing WLC devious seemingly this mastering for followingThe recommendationsand tips chapterssome document highlight will base MLC, SALC andzNALC, butthey are less widely used. aggregation).SysplexPSLC xWLC metric, machine by machine (or group of machines for customers eligible for an by specifically and mechanism, MLC the via chargedare al) et CICS, DB2, IMS, z/OS, (I.E. competitors direct no have which chapters, previous the in mentioned benefit. derive to customers for appropriate, seems imperative underlying Its nuances. subtle with complex very is bill (MLC) However, all is not as rosy as one might think, because this Monthly License Charge Mainframe configurations have «exploded» insize the recent years. since especially and licenses of amount the determined MIPS) (E.g. power appears to be more favorable than those prevailing before, billing software when of the type only this glance first machine At promise. beautiful a need»; you what pricing. «dependent» say might some as towards moved have customers Mainframe of majority the 2000s, the Since further analysis istheonly wayto make sensible andprudentdecisions. these maintain and develop skills. to struggle companies some as case, the always tariffs. associated and contracts these manage to skills specialized However,WorkingCRIP the membersof all Group agree detailedthat and Figure 4:IBMMainframe Software PricingTimeline T C (One C here are also other types of billing metric, including including metric, billing of types other also areThere ime Charge + S&S – Service & Support) and Support) & Service – S&S + Charge Time e B soa te ws Py ny for only «Pay was then slogan IBM The T o conclude, all of the IBM software IBM the of all conclude, o T pcly hs contract this ypically his is not is This

PAGE 29 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 30 otat a bcm vr oau ad usinbe epcal rgrig the real regarding benefitsitdelivers to theMainframe platform. In fact: especially questionable, and opaque very become may contract the ultimately but discount, significant a obtain to company the approachallows to the overall approach we will discuss in the next chapter. In theory, this bundled dependent MLC, O multi-year The by therespondents to oursurvey, andinfact, generally. framework is known as an e osrcin hs o this of phase construction The makers. message is not readily communicated«paradigm» or new easily understood this by customeralthough WebSphere,decision- of version Systems Distributed the with competes easily even and attractive seems customers (z196/z114) Server WebSphere. A new mechanism, IWP for WebSphere and latest generation zSeries last decade or so for Mainframe software costs, in particular for new software like the in innovateto attempt an continued has IBM recognizethat must we Finally, (caveat beware” “buyer thesaying has emptor) beenmore appropriate! Never al. et flexibility, threshold MLC endorsements,O with possible as certain be and outset the from possible as wisely as negotiate • TzO he ee s o ob ta te itiue Sses PA rdcs generate products IPLA Systems Distributed the that doubt no • T is here • The low watermark MSU (CPU usage threshold) MLC is difficult to fix because •  Generally in this situation, an IBM zO IBM an situation, this Generallyin useless arguably and restrictive a in consider to impossible almost the real beneficiary discount, even if the actual prices of zO of prices actual the if even discount, beneficiary real the the budgetinadvance except incase ofuncontrolable over consumption. determine to principle, in allows, it that is threshold watermark low this of set too low, the associated discount rate is not very attractive. if the resource amount is set too high, the customer may pay too much and if overall figure, and explicit product costs are notidentified. exactly how much such a product costs in these scenarios, as the uselessproducts the via a choose competitive zSeries to product from another be ISV, without adding “shelfware” should or desire the while product, Systems Distributed IBM mnmns ahr hn o hleg te nelig otat o fa of fear significant billincreases, for especially for theDistributed Systems software. contract underlying contract the challenge renegotiate to than to rather amendments prefer policymakers because an event, from exceptional withdraw to difficult is It zO IPLA/ of plethora the with compared when low, remains products eligible attractive.very seem T C products. T C (zSeries One (zSeries C S cnrc cnit o svrl ilbe lmns (Model elements billable several of consists contract ESSO T T C clarification O prices,C Sub-Capping, C, IPLA, Distributed Systems Software, et al) and corresponds owever, the simple fact remains, the number of MLCof number simpleremains,factHowever,the the ESSO (Enterprise Services and Software Option), used . ISV) productsCharge)ISV) becomeTime competitive (I.E. ESSO.Conversely, becomesit very know difficult to S cnrc i teeoe rtcl W must We critical. therefore is contract ESSO T S areet uls tee s an is there unless agreement, ESSO C product will be replacedanother be by productwill C T C metrics, exitclauses, metrics, C The only benefit T C via an via C ESSO is an ESSO. ESSO Benefits ofthisapproach offers. technical quality for each and every product, as we only can get what the supplier accept this kind of approach does not deliver the best software product in terms of performance of the software components is not the main focus. We must therefore buyer logic, it serves costs objectives. It is not a technician logic, so the technical the of size the from leveragecontractcreateto lever a formainly negotiations, financial. to is approach this of consequence second The a great dealofotherISV’s. softwareof fromproviders.need other the Inevitablylimits dismisses which this required,functions software the of part large a provide to able indeed areISV’s with an extensive software portfolio : BMC, CA or IBM. Arguably only these major The first consequence of this approach is to limit the choice to only three suppliers widest possible spectrumofrequirements. Such deliver. to Management, et al. supplier Data the Management, Performance Monitoring, Operations, for include will functions expected functionalities and components softwareof maximum requirementthe a placescustomerincludes the that end, include which software. of amounts, lot financial a of terms in contracts major with left is One Approach Ultimately the overall approach minimizes the number of Overall contracts and suppliers. The A. business system categories; IBM document, three this Systems, (Operating in software into discuss not down do we broken which software, application be can software Mainframe 3.1.2 ISVnegotiationstrategies: holistic/generic orpiecemeal/ad-hoc was the only situation where suppliers agreed to a contract independent of the of independent contract a to agreed suppliers where situation only the was MIPS/MSU usage for several years. Indeed, for many years this type of agreement A common reason for adopting this approach is to safeguard against any increased managing multiple contracts. or ISV’s of number the rationalizing either suppliers, these with dealings their ISV’s.niche or importance can come from different types of supplier; IBM itself, major global ISV’s and lesser previously discussed, and finally other software systems. •  • Tin commitmentcontract of durations long with negotiate to opportunity he •  •  Access to billing practices that are more flexible when compared with a with compared when piecemeal flexible (case-by-case) transaction. more are that practices billing to Access order to get,even more benefits. A compound impactofoverall discount, related to thevolume purchased. is stronger thanon acase-by-case basis. favorableterms. with discount they more the supplier, Arguably the most important lever for negotiations; the more you buy from a i apoc my ean ny n o to suppliers. two or one only retain may approach This Therefore the objective for each supplier is to encompass the T wo main strategies are available to the customer in customer the to availableare strategies main wo T ransaction Processing Monitors and Databases), and ProcessingMonitors ransaction These software products his is a commerciala is This he buyer’sposition The T this o

PAGE 31 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 32 from 12 to 18 months in advance. What to do? Stay on this type of agreement? of type this on Stay do? to What advance. in months 18 to 12 from contract this of termination of issue the again encounter will company the Then alternative. an have we occur, problems if that knowing software, ISV’s the for installed. We must also consider alternative techniques or replacement products MIPS/MSU of evolution any accordingto in cost potential of terms the negotiate of the contract, the question of what happens on the day of release (exit). We must Firstly in the contract, we must ask in advance from the supplier and upon signature This observation hasbothcontractual andtechnical aspects. such contracts, it is always important to define the end point, as soon as possible. imprisonment, or no room for maneuver. supplier, ISV. chosen same the from their positioning software to of lot relation a with with closely very move must company The because itfreezes the situation,butthefinancialflexibility isdiminished. edged, with pros and cons, as this type of contract protects in times of uncertainty, these cases until the end of the contract period. software. used the just not contract, the with associatedsoftware all for pay to continues company the to belonging software of removal subject. delicate a always contract phase, negotiation the the during platform, Mainframe the situation this for for plan must We adjustment. without completion, power until continues CPU in decrease a deliver optimizations If associatedsoftwarechanges. companywith situation cost,the if severalallowcontractualforyear«realThis agreementchanges not time» does Constraints &limits cost savings. and this flexibility more customer, the ISV, for and loyalty, the customer generates principle For portfolio. the from product alternative an select to “credit” software a its of 12 to access gives products. If supplier the customer replaces or eliminates the one of these products, they have with contract a example, For installed.software the of some change to scenario (M&A) Acquisition & Merger holistic logicalappropriatea potentialsoftware is allowingin fora interchange. it Again, the Indeed, ISV. ISV, an major from portfolio software a of a majority the to access of gives often approach portfolio make the to within needs software customer to the if changes model interesting an also is proposition This pending thetimeframe whenoperations once again become better controlled. known. capacity planning is difficult, as the growth of the new workload is not necessarily they inherit new platforms, and must consolidate software, while CPU (MIPS/MSU) activities. (M&A) Acquisition & Merger or merger company of because either resources, agreement is especially interesting in contexts for the consolidation of Mainframe adjustments. and roundnegotiations new of therea which was after years, three least at for costs software their upon impact not did of platforms growth its the that meant this company, the For capacity. Mainframe installed herefore this type of agreement provides a protective tunnel contract, tunnel protective a provides agreement of type this Therefore These consolidation activities generate challenges for the customer, as he customer does not benefit from their efforts in either of either in efforts their from benefit not does customer The trrs (as cnrc, hr the where contract, (mass) Enterprise Hence this conundrum, by entering into Therefore this practice is double- here is a risk of pseudo of risk a is There e ae ple for applies same The his type of type This numbers for processors, cores, hypervisors andvirtualmachinetasks, etal). certainlyand farless complexrequire forDistributedSystems(which than asset environment, Mainframe the in follow to difficult very not arecontracts the that solution need. the the fits best that choose should and approach driven cost. contract than and rather need product functional on based suppliers and solutions market best price. the for product, best the choose and Approach area, functional by benchmarks and tenders conduct to is breed) of (best approach known (case-by-case) well and traditional This Piecemeal The B. well before therenegotiation, benefittingfrom true relationship management. relations with the supplier, regularly working with them during the contract, and The last constraint regards what kind of contract is required to maintain very good change threat from theircontractingto software supplier.subject time any at be can company customer this but budget, little disappears. software the generally and arguablydeteriorates, relationship the if agreement, rental a updates,evenor maintenance-free totime allow In fordisengagement. of plan a continue using the software via a simple maintenance mode, without license code can they and company, the softwarewith staysthe supplier, fromseparatetheir depending on renewal or not. In the case of acquisition, if the company decides to agree on use rights, but at the end of the contract, software usage is in question, MSU, and publisher the tariff.updatedand buyeragreementan rental the and on In MIPS mode, of evolution the the to reassessment of a end is the there At contract software maintenance. of rate the and package software the of price purchase the covers agreement the mode, acquisition In use? to right the with tenant a simply or service, a pay I which forsoftware of owner the I am heavily, software. of rental or acquisition the for contracts covers agreement the whether ask must one approach,overall of situation a in Finally, risks, because thecontext ofenterprise computing changesrapidly. are there years, four Beyondstability. more generating maybe years, four over three years, which is a good pace for negotiations, but there are longer contracts, contract. the of length the assess to have also We return. no of point a to deteriorate relationships supplier the the ever consider if escape, to of months means 18 next the potentially and contract, this for platform over three years, where the company spent 18 months optimizing the Mainframe type contractthis ironicconsequencea of An the with scenariopenalties. canbe financial big and risk high associated with termination, contract anticipate not do preciselyusers becauseis it past, the reputationin bad a such had contracts these If choices. technical associated and change contractual the of consequences anticipate and deadline this for prepare must We competitors? to Migrate he objective here is based upon on the systematic review of competitive of reviewsystematic the on upon based is here objective The e es cnrc my e trcie o csoes with customers for attractive be may contract lease The he company agrees to have so many suppliers, knowing suppliers, many so agreescompanyhaveto The T oday, the typical duration is duration typical the oday, his question weighs question This i i a is This

PAGE 33 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 34 Disadvantages Advantages e o-ooeet o te utmr otae niomn generates environment software customer the of non-homogeneity • The •  •  •  •  •  •  •  • The abilityto contact allsuppliers (ISV’s), major,intermediate andniche. •  •  greater difficulty in integrating technology and a proliferation of ergonomics. that isproactive anddiligentto facilitate thesechanges. supplier a requires it more the environment, our into integrate technically subsystems. major its and z/OS vis-à-vis software, more have may of integration and interaction of type this of mindful be difficulties. must We ISV niche a whereas consistency, and currency maintain to resources the have generally ISV’s customer.Larger consistencythe for and challenges and currency versions generating new products, software of their delivery to the updates in behind fall ISV’s some Occasionally products. effort,remainand technicalof mindful consistency for interrelated software technical more requiring consistency, platform maintain to effort an make major one With partner, this update requirement products. is simpler. software With several suppliers, we must other of versions updated generating versions of Operating Systems and other major IBM Mainframe components, associated and z/OS with major subsystems. versionsFor in zSeries, softwaresoftware particularly companies of are rapidly consistency consistency, adopting new the technical to regard a greater maintaining in Difficulty technically thancontractually based. deals. piecemeal to opposed as volumes, included inthecontract. the customer expended one or two years effort to implement all the software wherecontract,Mass) or (Enterprise packaged complexitya comparedwith again. once process the selection enters software just piecemeal but products), software of tens (maybe portfolio software ees o cmeca ngtain r salr hn oprd ih a with compared when smaller deal. packaged are negotiation commercial for Levers Obligation (duty-of-care) to manage relationships with many suppliers (ISV’s). supplier.the with business, or problems,technicalcontractual of case in System Information overall Mainframe the risks for significant concernsor generic globalor No company ceases to usesoftware, they stop payingfor the software rental orsupport. if rationalization; software during reduction cost Immediate Units, Value (E.g. model other Number ofresources, etal),optimizedasperusual. any on or partition, the for license, MSU a a for apply can In approach piecemeal a whereasimprecise, so software.is which model, of type or licensing group uniform a is thereapproach, Mass) each or comprehensive(Enterprise for model contractual the Adapts Functional coverage for theproducts best suited to customer requirements. and willnotforget itthedaywhenyou needthem. out it realize quickly will competitors as ISV, same the from buying always adjust their prices. Be careful though not to practice false competition by not Benefiting from continued competition between suppliers that drives them to he customer will prefer to negotiate deals based on large on based deals negotiate to prefer will customer The he company is not challenged with replacing an entire an replacing with challenged not is company The herefore negotiations are more are negotiations Therefore he more software we software more The i mto hs less has method This maintenance. of revenueout get will ISV the as free,forsoftware the get even sometimes and customer new ISV. each the for that revenue additional means brings which base, user installed the by supported areevolution and maintenance of manufacture.Costs to nothing costssoftware developed, once Indeed, model. financial its of because bargain to scope ample provideshardware, than rather software, that recall please introduction, the In 3.1.3 ISVcontractual arrangements: licensing modes&billing software additionswithsmaller nicheas opposedto larger ISV’s. get to better is It they’representedwith. opportunity limited the maximize to try because flexibility, demonstrate areambitions their will necessarilynot unreasonable, whereas rivalbigger will a ISV ISV niche smaller A approach. their stiffenbusiness to them leading competitors, their of one (Enterprise with comprehensive approach a mass) or in engaged already company volume low a a with do business only will of they know will ISV An additions. software these for nothing is which embarrassing suppliers,in itself, but requires our us to manage be very careful and with our attitude with to ISV’s, engage must We inevitable. producttheir unsuitableis unusable.or diversificationSome ISV’s of therefore is or portfolio, its productin a missing is requirementsor system,Mainframe a for This overall approach has obvious limitations. No ISV covers the entire functional unfortunate for boththecontractual process andthetechnical approach. of day renewal the replacementor arrives.fact, becomingIn is attitude scarce,this it’sand when for products, and ISV’s emerging detect helps avoids, testing a good knowledge of the software market for a few major components, by regular renewingsoftwareareor weincumbent terminating product.an not Maintaining a continuousand have when process of creationeven andProof-of-Concept (PoC), Another good practice, but unfortunately too rare; regret, we must gothere from timeto time. year.per With €) or ($ million severalare savings the sometimes but activities, technical and training for days project Management, Change requiredfor effort and accept the effort of a migration job. Such a decision is costly, with significant process, the with not, or radical whether harmful, if throughrelationships,even break ISV to howlearning go even we to Sometimes able necessary. but if ISV, them incumbent replace an with engage to bound duty not is customer the that clearlynoted be regularshould Practicing competitive tendering withISV’sit ; there are additionalcommon practices to consider: conclusiona As to chapter,this beyond choicethe betweentwo these strategies, from asingle ISV,to deliver functionalortechnical uniformity. applications packaged of groups small consider to mindful be always must i i epcal tu fr ytm Mngmn sfwr operation. software Management suppliers. of diversity a deploy we Systems for true Sometimes we deprive ourselves of technical and functional synergies when especially is This here is ample opportunity to negotiate, to opportunity ample is There herefore, with this type of approach, we approach, of type Therefore,this with we shouldregularlywe test software

PAGE 35 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 36 1. The IBM usage based model, with a principle of billing the customer, measuring W business usage). with compared growth CPU (E.g. usage have software may with increase connection MSU no the because also but cost, entry high the of because disadvantageous MSU, installed of method billing classic the than paths other position. Let us be creative and innovate on how to be licensed and billed,over time, while exploringavoiding making gestures with ISV’s that this is not a sustainable We must be innovative with our suppliers, hence the importance of continued work appear. management and monitoring of difficulties the and explode, costs machines, of the entrycost islow, butifthesoftware isthendeployed onasignificant number this erase Distributedcoursedrawback;thereSystems tothe world, of a effect. Becausein is ways devise to is creativity customer associated and negotiation canstart inexpensively singlea on server lowwith power. environment Systems Distributed a of deployment the whereas mass, initial of effect this fromsuffersplatform zSeries the So acquisition. platform of cost the comparescompany software,installingthe When platform.SystemsDistributed generating metric. resources,mass initial MSU an and MIPS global by measured typically platform, environments nature. (E.g. centralized its of because penalized often is Let’s summarize the issues associated with software licenses; the zSeries platform codes). (E.g. prohibited or restricted is use the contracts. Some ISV’s consider that without maintenance on payments, softwaredepends always It maintenance. for termination, of case in including rights, (T cost increasing of Whichever becausemode is chosen, you should check that appeal the contract has perpetual use its loses mode rental the time Over routeacquisition moreas lowerattractivethe renewal becauseof support rates. Should I buy or rent? Regarding software, the general tendency is to consider the the Mainframe. on (HPC) applications computing technical and scientific processed companies zSeries the on platform. run Particularly applications not that use a lot of should CPU cycles. workload certain obvious; the remember Then e can distinguish four models: the environment (E.g. environment the of subset a is which MIPS, or MSUs of volume a committedto customeris the the component activity where the invoice is proportional to the activity. Initially, the advantage of avoiding invoice payment as per the installed MSU or MIPS. or MSU installed the per as payment invoice avoiding of advantage the technically.fail components certain of activity the measuring measurement toolthis is bundled cases, with the software.some Sometimes the possibilitiesIn of al? et often, how how, when, usage, actual the measure consume. will software the what The Mainframe CPU cycle is expensive, so this is not a good idea. T est, Production, et al.) generally comprise a single Mainframe his is often the number one argument in favor of the of favor in argument one number the often is This T est, Production, et al.) MSU established by estimating by established MSU al.) et Production, est, hen the customer and ISV agree on how to how on agree ISV and customer the Then hey will not issue new CPU based license based CPU new issue not will They eeoe l Mainframe all Therefore he wholeThe ISV of game Historically, some his model has model This CO). 4. Texplored,be should rarelyablethat now being one method, last but he used, 3.  2.  business locations. of number the year, of per transactions number CICS of the number the upon users, concurrent based charged software technical some encountered study our exampleFor agenda. the on is agreement of activity type this their business, on of impact the about care increasingly managers Operations IT as But culture. computer CPU) (I.E. traditional with fixed still are who ISV’s, creative. and business or associated usage activity, as a way to measure, trace and audit withthis unit. product We must be imaginative associated of work unit the find constant. remained service the since costs, increased no user.the protect to is increase, the software bill will also, which is perfectly legitimate since the goal MIPS). per not user, per billing software, security (E.g. negotiations their during method this about think not indicators. technical Mainframe to opposed as usage) (business units work on based models billing derive to the of totality Mainframe platform. the Thismethodistypically usedfor packagedapplications. upon based payment license software avoiding mergers, business of case the in preferredtechnique a also is It nature. by transverse more software, technical for than software application enterprise for better charged in proportion to that partition. be only will partition isolated an avoidon installedapplication An costs. to globalizing functions separating partitioning, by perform to is method Another fusion- transient and grouping contexts. activities development example, For assessment. of determine the use of said database? It is then necessary to find other methods we to how capacity,environment the of 5% uses only which application, given a with associated database a of case the in but complicated, not is partitions T (E.g. environment entire the into embedded or integrated more functions for functions and activities classified as isolated, becoming a source of controversy 2,000 MIPS. its Development environment is 2,000 MIPS, proposing an invoice based on the believescustomer the capacity, MIPS 20,000 of Mainframe a installedon tool, development a of case the in example,Foragreement. contractual the revisit absencemeasurementthe MSU In of tools for whatever technical reasons, we Compuware for Development T by practiced is method billing this example, completely For results. not desired did the achieve but hopes, many raised has method billing this time, its For the customer, this avoids the requirement to pay based on capacity. During est, Production, et al.). In the case of Development, assessing the capacity of Development,capacityassessing of the case the In al.). Production,est,et This is known as confinement contract. hese are rarely used areas of discussion and negotiation with negotiation and discussion of areas used rarelyare These owever if there is an increase in MIPS, there will be will there MIPS, in increase an is there if However ools software products. herefore, if in the future the number of users of Therefore,number futurethe the in if his approach is unfamiliar to users who do who users to unfamiliar is approach This This very practical solution works much This model applies well to he challenge is to is challenge The

PAGE 37 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 38 3.1.4 ML determine thepower consumption andbillable. speaking). (relatively time over stable measurement hardware,of this generationkeeping new each with changes MIPS/MSU ratio machines. two these on MSU of number same the exactly consume machineof will ona run z196. latest to the on minutes 5 in run hour will maybe an period, time that took in1990, that a program Specifically, in onehour,today itisameasurement unitdefinedbyIBM.) processingperformcomputercanof a amount (Yearswork the was MSU ago, task. a perform to consumedPower’ ‘Computing measurementof a is MSU a The MIPS measurement is a unit directly related to processor power, whereas What isthedifference between MIPSandMSU? •  With regard to billing,z/OSsoftware can beseparated into two large groups. General 3.1.4.1 TheIBMbillingmethod •  • CPU consumption (MSU/MIP S) at defined time intervals that can be used for where thesoftware runs,including: (LPAR) partition the on information usage recordscontinuously system z/OS activity. partition associated upon based is payable amount the product, products, eligible VWLC on exclusivelywhich typically constitute the focus bulk of the high-cost software. we For each software sections, following the In MLC). as Flat (known cost fixed at leased products • T also are here •  (R4HA MSU). time is recorded as well as the moving (rolling) average over the last 4 hours MSU consumption (Millions of Capacity Planning. sfwr aqiiin ehd ae zO named method acquisition software A be saidabouttheminthiswhite paper. will more nothing amount: invoice total the of 3-5% generallyare products VWLC (Variable Workload License Charge). Charge) usage based License tariff for eligible (Monthly are products MLCsoftware these of a Most metric. on based rented; be can that Software than thezSeriesserver capacity. the company can invest proportionally, based on actual consumption, rather that meaning sub-capacity, for eligible generally is product of type this as negotiating the purchase of software based on the detail. in real discuss not needs will of the enterprise, C: ontrolling thebillinglevel he only important thing to remember is to start to is remember to thing important only The Service Units) per hour. T (One C eeoe t s sd to used is it Therefore m Cag) ht we that Charge) Time The average elapsed owever, it However, These The The The following formula summarizesthismethod: the processing partitions software.all for R4HA the of sum the by month billing the in reached value average of MSU consumption for each partition. where 4 server(Rolling R4HA each the denominates on It processes. it partitions all consider we product, software VWLC eligible each For Calculation mode the timewhenthissumreaches itshighest level duringabillingmonth. consumption. usage average the of sum the of peak the for (CPC) machine each for established is level billing monthly the billing, usage VWLC With Metrics usedinVWL monitor themaximumusageofR4 HA metricsfor several partitiongroups, to reduce cost. Therefore, ideally maybe this software is not installed on all partitions, and instead we should higher. processing,overnightbatch the thus during and R4HA a on based be will billing the them), of that reaches its maximum consumption during these batch periods (and it is the case for many processing.machine batchYet, a during overnighton usage minimal has and day, the during situations. paradoxical to lead canCICS the example,For which difference, essential an consumption, own its to wherepartitions consumptionprocessedof the Softwarebilledon accordingbased is not and Observations aggregation mode,theircombination isusedto determine theoverall software invoice. (CPC). (Sysplex)PSLC in server if or server, per invoicesoftware create zSeriesa to used then areobtained physicalvalues each for performed then is calculation This product. software associated the for points usage MSU maximum more many be will thereproducts, software many for tothe So product. software one for (graph), figure above corresponds the in capacity) curve green the installed MSU (970 figure 3. Period in partitions usage all by consumed amount MSU maximum MSU 900 the case, this In each LPAR, which soproves farmore to advantageous thecustomer. NB: Thisisthemaximumcombined amount for allR4HAmetrics, andnot thesumof for themaximumMSUusageachieved T Figure 6:VWLC Rolling4HourAverage (R4HA)Example P) monitor has very reasonable CPU usage CPU reasonablevery has monitor Processing(TP) ransaction C billing Figure 5:VWLC Rolling4HourAverage (R4HA)Formula The invoice is established upon the maximum our Average) which is the moving 4 moving the is which Average) Hour he period (time interval) chosen is chosen interval) (time period The his examplecorrespondsto This e oa MSU total The Hour

PAGE 39 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 40 z/OS V1BaseCost E MSU: Highest level MSU: Lowest level Product Name data and tools to produce reliable cost figures. is therefore essential to ask IBM to simulate the cost evolution. In any economic simulation of this type, we must be suspicious of any abuse and linear costs. It only in since applies to theincrementalanother, MSU metric,whichisalways alower cost perMSU. to Mainframe one from application an porting when savings cost the costs operating minimizing of effect the related to increased has MSU power (E.g. data center metric consolidation). Conversely, reductionit also minimizes cost MSU incremental This customer always paysthe€4,146cost, for 1,2or3MSU. products. Please note that the upper and lower thresholds for each level (MSU) are common for all VWLC An example ofz/OSpricinginstalled onaz196(AWLC) inFrance: installed capacity. There is a ~10:1 ratio between the highest (10) and lowest (1) incremental MSU cost related to the business prioritiesdictate thechoice. foreverwas recognized.thus, It be must objectives availability continuouswhere Sysplexand focus some software on a limited number of partitions. But this may be a paradoxical approach,We verycloselocalizationpay the toattentionsoftware,should of whereadvantageous is to it in thefollowing chapters, willhave adirect orindirect influence onsoftware billing. described invoice,controllingas software(SCRT) of the methods and tools of use Finally,the will follow, soperhapsagroup ofpartitionsfor DB2,IMS,etal,mightbeabetter option. usage al et CICS, DB2, then usage, z/OS base and underlying the on partitions all base we If However,pricesfromvaryproductunit one another.to Figure 7:MSUPricingStructure –DecliningCost for AdditionalIncremental MSUUsage 4,146 Base Figure 8:z/OSviaz196&AWLC Incremental MSUPricingExample 3 1 Level 385 45 1 4 Level 315 175 46 2 Level 226 315 176 3 Level 120 575 316 4 Level 875 576 They alone have the necessary 92 5 he first the fixed,The so is cost Level 1315 876 64 6 Level 1975 1316 48 7 Level 1976 MAX 38 8 Below isanexample graph showing the“billing”month,accumulated for eachpartition. month associated withthe“highest amount”isusedfor billingpurposes. month, the values used for the “MSU Billing” of each partition are added. valuesareThese previously.calculateddescribed methods the by billing the Forin hour each each server and for each eligible VWLC product, , which acts as a basis for software invoicing. Each month, the customer provides IBM with a summary of the usage (MSU) consumption for The monthly invoice calculation basedonSCRTreports usage consumption CPU as perthedistribution ofsoftware products onallpartitionsandservers. measure toR4HA associated the of evolution the requirementas well as (CPS), servers zSeries physical for a (metrics) demonstrates review previous The Measurement activity. tools, whereas more sophisticated tools are required if the peak is originated from Peak load activity related to batch processing can be reduced by use of simple and rudimentary the choice oftools to implement optimization: The nature of the workload activity (Batch or On-Line/TP) during peak load times will determine available options,asfollows. computerforoperationresourcesall of available frommarket, reviewwhilethe will we some principle the is availableresources,this the and of fromall most the make to is solution The for applied rarely is control operations transactional activity becauseon zSeries servers, and problematicmust always deliver service more commitments (SLA). even is operation This We ponder and perhaps still use these methods in marginal and the most extreme of situations. constraints. unavoidablebusinesscertain, for generate would and productionplanning, reviewof general batch (E.g. to related time primarily is over load workloads peak activity. the of if solution, distribution possible a better scheduler); batch a the perform via to imagine could one Ideally, seems compelling for two reasons to eradicate thesepeakloads. invoice (operating costs), but also on the sizing of servers purchased (investment). It therefore As per the previous commentary, peak workloads not only have a direct influence on software 3.1.4.2 Theproblem of smoothing (consolidating) peakloads based onthe“MSUBilling”calculated for 28July, from 12:00-13:00,at767MSU. AWLC, (E.g. xWLC for eligible product A partitions. all fromcalculations billing for referencetime the be In this example, the highest value is for 28 July, from 12:00-13:00, at 767 MSU. It will therefore However, such an implementation can be long and tedious, because it would impose a WLC, VWLC, et al) billing running on all partitions of the CPC will be charged be will CPC the of partitions all on running billing al) et VWLC, EWLC, Figure 9:Example Multiple LPAR MSUUsagefor SCRTReportSubmission The peak time in the TP (on-line)

PAGE 41 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 42 reserved to highly skilled specialists. invoice.softwareresulting the thus consumption, and overall reduce to algorithm simplistic a is It value. soft-capping a assign and LPAR in the LPAR group) group everyto applies MSU forGCL (E.g. isomorphic an define to is mechanism this each use to way basic A zSeries server (I.E. containing all of the server partitions) associated PR/SMweight setting. via implemented is MSU, as expressed value, This for account into taken thus is metric activation. Soft-Capping after new billing WLCcalculating A value. chosen arbitrarily an via limit partition any for period hour 4 a during consumption MSU smooth topossible been has it 2004, Since Soft-Capping •AutoSoftCapping (ASC)software –Appliedto agroup ofpartitions(LPAR) orserver (CPC). • • Here we describethe3primaryexisting systems andtools, from simple to more advanced: 3.1.4.3 Themethods andcontrols for controlling thesoftware invoice architectures. It is now possible to define groups of partitions that are assigned limited 1.8 are assigned that z/OS and of partitions z9 groups the define from(MSU). values consumption to emerged possible now design is soft-capping It original architectures. the from evolution An Soft-Capping Adaptation: Group CapacityLimit(GCL) but notfor overall management(E.g.manyLPAR’s, 1CPCormany CPC’s). (LPAR), partition specific a for adapted metric, manual and static a Thereforeis soft-capping the DC,whereas theR4HAreached 450 MSU. In the graph below (Figure 11), the Billing amount (purple line) shows a level of 380 MSU as per smallest the but MSU amount,eitherR4HAorDC: R4HA, the retain systematically not does calculation SCRT the Therefore, system limits(Caps)instantaneous consumption IMSU(red line)attheDClevel. From(greenbelowgraphthe R4HA exceeds line), the (Figurethe line) (blue when DC 10), the > Therule ofSoft-Capping: sum ofDCMSUtotals represents thelimitofMSUfor software billing. This dictates the MSU billing limit we wish not to be exceeded for the partition. from thegraph below (Figure 11)represents aSoft-Cappingexample. Group CapacityLimit(GCL)–Appliedto apartition(LPAR) group. Soft-Capping (DC:DefinedCapacity)–Appliedatthepartition(LPAR) level. Figure 10:SCRTsoft-capping calculation formula –R4HAorDC Figure 11:Workload soft-capping example –R4HAvs. DC ach partition within the group utilizing MSU resource, as per its resource,per MSU groupas utilizing the within partition Each owever, the daily usage of this algorithm is strictly is algorithm this of usage daily the However, MC at the partition level. partition the at HMC his is the “Defined Capacity” or “DC”. or Capacity” “Defined the is This The aggregated he blue line blue The • • • AutoSoftCapping allows for: activity information inaz/OScomplex. T Directors, Managers, (I.E. usersdifferent gives interfacegraphical and reporting the Finally, on a on apartition-by-partitionbasis. based profiles activity differentevery month, of end (E.g. manage period specific dynamically also can solution software The allowspartition, safeguarding atalltimesthemaximumavailability. each algorithm for requirements MSU sophisticated the considering sharing, A MSU for flexibility MSU and partitions. control greater by sharing Mainframe assigned available, Capacity” the “Defined between the resources optimize to is principle underlying The that supportssoft-capping. provides Management) zCost environmentany words,for other in onwards),or and (1.7 versionregardless z196) z/OS and the of (from z10, z9, (z900, servers product of models many for software soft-capping, Mainframe IBM for AutoSoftCappingflexibility the 2006, Since AutoSoftCapping: anadvanced software tool business deadlines and SLA metrics (E.g. batch processing, transaction response times, in et delay associatedal). and service of quality the degrading of risk the courseL) of is disadvantage GC workloads.( customer for ~10% by utilization Limit MSU decreases practice in and apacity C roup G moredeliveringpredictable of advantage the softwareMLCamounts, has based method This sing U 1: ase C commentary. Wespecific willnow focus onalternative methodsofMSUoptimization,asfollows.any need not does soft-capping basic such, as practice, general a as accepted Now feedback andrecommendations3.1.4.4 Survey • •The AutoSoftCapping product has increased automation and dynamic optimization benefits, are features (GCL)) Limit Capacity & Group Capacity (Defined soft-capping basic •The Elements ofcomparison between thesesoft-capping tools cncl Oeain, t l sae vsblt ad n nesadbe ee o soft-capping of level understandable an and visibility shared al) et Operations, echnical, ae o sml ad nesadbe rps n tbe, o al eeat Mainframe relevant all for tables, and graphs understandable personnel, introducing acommon languagefor theentire company. and simple on based A simple and clear vision, with real-time behavior scores and billing level, with “reporting” better performance. This helpsto keep itbelow thelevel required withsimple IBMsoft-capping, while achieving Billing control, the Defined Capacity MSU total is permanentlyquota itdoesnotneed. controlled by AutoSoftCapping. not reach the MSU ceiling does which it partition is entitled a to when use, it AutoSoftCapping, allows other With partitions to metric. use the soft-capping MSU Capacity) (Defined IBM basic the by introduced limit fixed the than penalizing less much becomes It priority. and variableis algorithm AutodictatedSoftCapping , needs partition MSU by the on depending MSU resource optimizationandallocation between partitions.Thelimitintroduced bythis o ey pca css epcal ascae wt cnrcul biain, i--i some vis-à-vis obligations, software licensing contractual models. with associated limited especially generally cases, is special use very Its to . consumption MSU average) the not (and instantaneous the caps which mechanism restrictivemore far a capping, hard ancient the concurrentlywith such manage to requiredproducts capping soft these investmentuse to not advise usuallypractice good particular, human In systems. and budget, objectives, their to according simultaneously. tools these soft-capping of one that more use to inefficient is it viewpoint, cons and pros productive a From this software product isnotfree. reporting. management and controlreal-time for dashboardsprovides also and weight parameter. the from accessible cost. additional intervention/command) and operational no per (as dynamic products is these of configuration at available architecture, hardware IBM the within incorporated C O (ige bet prto) oe fr xml, h PR/SM the example, for mode, Operation) Object (Single SOO HMC eyn ms slc ter rfre mechanism, preferred their select must Everyone T uesday, et al), adapting soft-capping requirementssoft-capping adapting al), et uesday, he significant The However, The

PAGE 43 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 44 more ofalaborious operation. somewhat is records,directlywhich SMF fromqueries build also canWemeasurement. this for used be can Monitors) CMF, RMF, (E.g. measurementtools systemsStandard Mainframe R4HA. daily the occurs, monitor andto whatextent. it when to identify to placeimplementation, its during in impact Limit toolsCapacity Group of measurement have to essential is it Finally, level ofcost-control, butthequestion ofperformance still remains. with obligations contractual certain respect to certain meet software vendors, to including IBM). necessary is it when example, for exceptions, Therefore becomes partitions soft-cappingspecific irrelevant on these under (with conditions policy for allserver partitions. effective way to use the Group Capacity Limit (GCL) typically exists by defining a GCL (capping) the system to make trade-offs in times of constraint. Once these conditions are met, the most for priority sufficient of be must “Discretionary” as based classified workloadsscore, Similarly WLM a constrained. upon not are partitions Production that safeguard to consumption, low-priority for weights partition workloads PR/SM (E.g. Development, the scenarios, constraint global for particular, In realistic prioritiesandto influence the level ofMSUmanagement for eachpartition. T fact artificially capping theserver inasituationofduress orhigh-workload. penalized. be will workload which knowing of possibility no with available, globallyCPU controls the function (GCL) Limit GroupCapacity demonstrated,the have we As Conditions for success: pre-requisites without penalizingperformance (business processing).optimization, cost for setting parameter MSU GCL the control to is challenge the Therefore •The soft-capping level for a server (CPC) may be variable, ranging between a minimum and ) SC (A • apping C oft S Auto The ASCsoftware solutiondiffers from GCLasperthefollowing: sing U 2: ase C soft-capping with familiar become analysis demonstrates thecost reduction viabilityofsuch an approach.further to if cost, costadditional for tools, additional sophisticated more with without evolve then can and We techniques. way simple a also is It small proportions (butstill worthwhile), andsogenerally painless. This method demonstrates the possibility of lowering the MLC software bill, albeit in relatively on thecustomer workload). previouslyprecautions outlined can basic safely fewmake a minimum the MSU gain of via ~5% (sometimes moreimplementation depending careful and Diligent features. interesting has Without doubt, GCL is constrained by its lack of flexibility resource and dynamism, but its basic function MSU the reduces typically (GCL) Limit Capacity requirement ofzSeriesservers. Group of implementation The Observations &Recommendations equal to thetypical monthly peakload, gradually declininginsteps of5MSU. For example, we can start GCL implementation during a period of normal activity, with a value us from anyadverse consequences. with careful monitoring. Our onlyexperience shows that the if process,we follow implementationthis careful GCL route, it this safeguards implementation for method is by model taking small predictivesuccessive reliablesteps, which can no be easily is supplemented there Since Implementation o avoido disaster, thereforeis it mandatory toeffective setting havepolicy management a and maximum value, contrary to GCL, which accepts only a fixed value. fixed a only accepts which GCL, to contrary value, maximum not according to itsPR/SMweight). MSU distribution is based on the immediate (instantaneous) needs of a partitions score (and T est, et al) must be sufficiently low and activity below their usual Therefore this approach provides an extra his allows for more for allows This his function is in is function This i alw the allows This he objectives and definitions for soft-capping partitions and servers are managed and aremanaged and servers partitions soft-capping for definitions and •T objectives he a with time, real in situation soft-capping the visualize to interface web a has •T tool his o ucntand oe ol aheal wt AC blw Fgr 1 & iue 3 i an is 13) Figure & 12 (Figure below ASC, implementation example. with achievable only mode, unconstrained For ~12-15% MSUreduction. The gains achieved with ASC exceeded those achieved with GCL, in the same context, reaching • • With theASCtool, two modesofsoft-capping managementare possible: • • • •The tool supportsmultiple Mainframe servers from oneunified focal point. (3) (2) (1)The maximumR4 HA value2,777 M isSU,asper01October at 12:15. This graphical analysis ischaracterized bythethree following points: without MSU ~3,500 soft-capping,of from 30September-02 October 2010(peakbillingperiodfor Septembercapacity 2010). a with server z10 a of behavior the shows below graph The activity workload profile, butit remains achievable, withoutimpactingperformance. partitions). to constraintapplied preferredis (no (QoS) Service of Quality the mode, activity.workloadthis maximum In during bill, softwareWLC associated the and usage, MSU optimize to is objective mode’smain This Unconstrained Soft-Capping mlmne b te otae nt via (not software the by implemented standard range of36hours (maximum96hours). significant gains to bedelivered. performance). reducingassociatedpolicy, per cost, safeguardingas soft-capping, Constrained (Optimized usage). MSU Unconstrainedhigh of associatedperiods per soft-capping,during as (Minimal policy, It MSU. potential identifying supplier, the from available is contributes to aparticularimplementation ofsoft-capping, calm andcontrolled. tool simulation a and Finally, logs, the via generated of exception situations. be can alerts encountered, communicatedby scenarios the on Depending via aschedule. Profiles (policies), as per customer requirements, can be defined and activated automatically profile). rcs wl b etbihd at established be will process2,754 MSUasperthe01 October 12:00time period. the SCRT by calculated level, billing the profile, this For MSU consumption (IMSU) falls quickly and sharply from 00:00 (end on-line day/TP activity). Figure 12:Unconstrained workload example –R4HAvs. SCRT&IMSUanalysis mail, SMS, et al, safeguardingresponsivenessal, management better et and SMS, Email, he MSU reduction will therefore be minimized for this high this fortherefore minimized reductionwill be MSU The C ht eurs ces ihs i a particular a via rights access requires that HMC

PAGE 45 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 46 xettos n te otata cmimns soitd ih oe ao software major some with associated commitments the contractual meet suppliers. and levels, the consumption (CPU) and MSU expectations master to required are tools such Finally, This advantageous software tool enables arapid implementation ofcontrolledbill. soft-capping. software (SCRT) WLC to the lower constraints, to implementation savings, performance MSU realistic and and/or run pragmatic develop finally benefits to any identify possible parameters, partition is and associated the server it refine to periods, ASC, time previous from with soft-capping of availablescenarios (simulation) tools simulation the Through Operations), andfinally, a strong shared commitment byallinthe company.(T environments administer who those with consultation and analysis, implementationrequiresdeployed,soft-capping software is extensive or techniqueWhatever those associated withGCL,delivering biggerbenefits. to options alternative using consider to us allows tool, ASC the processedwith Soft-capping, Observations &Recommendations the of performance) (reduced expense service. the at is gains, reduction MSU additional obtaining period. time 11:00-12:00 October mode to reduce MSU consumption. maximum activity peaked for 4 consecutive hours, it will be necessary to switch to constrained the suitable. If is machine profile the activity of the possibleonlyif is scenario this course, Of (3) (2) (1) behavior (Figure 13): thresholds, ASC and the MSU 2,000 setting at By abilities. optimization soft-capping variable and workloadthis profile,optimize dynamic and its upon to due act enableto will us tool ASC The represents a209MSUor~7.5%decrease for thebillinglevel. to reduces level billing the R4HA, highest the with period the during DC the lowers that change this With have noimpact(capping withoutconstraint). but their level of consumption (IMSU) remained below the associated DC, this capping will With the implementation of soft-capping (DC), some partitions are capped (I.E. DC

PAGE 47 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 48 optimized IBMMLC software costs, regardless. thus. forever was additional costs for and the customer to consider, as per normal, while metric, ASC software-capping has MIPS installed the upon based are costs softwareproducts,ISV whose by offset somewhat softwarebe invoicecanSCRT IBM the with Finally, when activating the additional engines (CPU’s), the savings and optimizations associated (4)The capacity upgrade (addition of a processor on both servers) was forecasted for November by regulated be will n, month for consumption CPU the to linked invoice (3)T software he 710 detailing the specific MSU peaks for related z/OS software products as per “PRODUC per MAX CONTRIBU as products software z/OS related for peaks MSU specific the detailing 710 Below (Figure 15) is an extract from an SCRT report dated October 2010 for a z10 Server Model SCRT generation process captures allofthesevalues (product peakMSUusageamounts). also are R4HA per as recorded, peaks almost certainly monthly different from partitions, that of the machine (I.E. z/OS MSU usage value). some on present only software, other For with associated metric billing peak R4HA. the on mainly focused far so discussed options Saving increases. For low MSUusagecustomers, larger theMSUunit price isatitshighest. for suited better is model billing (WLC)customers. MLC IBM the 3.1.4, section in explained As Options 3.1.4.5 OtherCostSaving • • • In terms ofsoftware tools, we must remember that: several for prioritization consumption, MSU of workloads (E.g.Production vs. level Non-Production, etal). overall as such objectives, technical assistance. no of being counter-productive,prove will consequently and imagined be can which problems,complex neverdesirableis implement tosafeguardsit simple particular, rules.addresstoIn intended However, it is always best to leave the system the maximum amount of freedom, to implement as pertherequired discretionary policies, asandwhenrequired. At partition or Sysplex level, a WLM policy that has sufficientAn effective andclearly flexibility differentiated server (MSU widePR/SMweight managementpolicy. resource allocation) to resolve anycontention scenario. Inthisinstance, itisessential to have: However,toprior deploying soft-cappingany approach, systemwenotethe that mustable be the most suitable methods aspertheirown individualrequirements. Obviously not all cases are suitable for all organizations. It is up to each company to determine Following this study, we have highlighted a number of mechanisms that generate cost savings. Soft-Capping ummary to purchase. free. not is it cost, is methods, previoussoft-capping the unlikeconsideration, only features.Its to MSU usage without soft-capping. It is more dynamic, automated, offering many interesting AutoSoftCapping (zCost Management) typically delivers a ~12-15% MSU reduction compared to implement, itsfunctionandflexibility islimited. save ~5%-10% MSU. Its disadvantage is will its static nature GCL (not dynamic), and although simple optimization, without and precautions specific without Used policy.group soft-cappingpartition a implementing for tool basic and simple a is (GCL) Limit Capacity Group as a tool for a comprehensive soft-capping policy available generally (E.g. Defined Capacity). Partition level soft-capping is easy to implement and simple to understand. It should be seen 2010. January 2011). November 2010 will be regulated in January 2011 (I.E. 2 month period - November 2010 to for increased consumption CPU example, particular this in So months). (2 metric n+2 an This method relates to MLC software, such as z/OS, active on all partitions of the server. Therefore a technical and financial benefits study is therefore required before deciding e S ui pie erae gauly s h nme o MU consumed MSU of number the as gradually decreases price unit MSU The ORS”: he rules (policies) put in place must be good enough to translate to enough good be must place in put (policies) rules The e wl generate will They The T • • •The combination of the IMS and bill CICS software products softwareon some partitions lowers the peak the lower to implemented without impactingperformance: actions of set a are below analysis, this Following total software bill. representingMSU, the 2,197 of was 24% which R4HA z/OS the than totalsoftwarehigher bill, representingMSU, However,1,473 of example,R4HA this a in has IMS that the see weof 27% as interesting as other software products, such as IMS, which only resides in some partitions. If we only consider the peak machine (z/OS) usage with a R4HA of 2,197 MSU, this might not be z10 Model 717servers deploying theVWLC* pricingmetricaggregated inSysplexfor2 (coupled) mode: invoice software monthly the analyzing of example graphical a 16) (Figure Below software invoice to furtherdetermine optimizationactions. resulting the for figures usage MSU these analyze individually can we information Fromthis uig h SC te let as ny h hget osmto fo bt o te installed the of both from consumption versions highestcostsnew)and and (old for newversion.the For example, let’s the consider V9 IMS an only Indeed, pays rules. client billing the holiday) (software SVC, SVC the IBM during the of view in savings some for allows Carefully choosing the order and timing of product upgrades/migration to on-line (TP)peaks(overalldifferent impactoflowering theR4HAfor thesoftware bill). partitions, moved the overnight R4HA peak (batch activity profile) to correspond with at least one of the Stopping the CICS and/or IMS overnight for the Development and Pre-Production partitions R4HA for only theseproducts. Figure 15:SCRTextract –peakMSUusagefor z/OSandrelated software products Figure 16:VWLC SCRTMSUusageanalysis for individualz/OSproducts

PAGE 49 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 50 3.2 Infrastructure Cost Reduction his approach is sometimes followed, usually by small and medium sized organizations. sized medium and small by usually followed, sometimes is approach This 3rd partyhosting (Outsourcing) induce they used, widely more major changesfrom atechnical, becoming financialandhuman viewpoint. are methods these and Although taken approaches objectives. different associated the detailed have we extreme, most the to simplest the From generating new opportunitiesfor infrastructure consolidation. generation of servers increases the architecture and performance when compared with the prior, commonplace.sector. al.) et Insurance, Banking, been (E.g. financial the noticeablein particularly is trend has consolidation infrastructure towards trend a years, several For 3.2.1 Grouping andSharingInfrastructure is asfollows: The than more by member Group Working CRIP 20,000 MIPS. bill the lower to helped 2008, in environment detailedServerpreviously,actions zSeries of combinationa The rationalizingassociated with • • NB. Abaseof100ischosenfor easyrepresentation (baseof100=100%). partition, where theIMS,CICS,MQandWebSphere software products are required. associated and workload MSU (TP) on-line the with associated not is partitions associated workloads. batch or (TP) on-line the for As part of a Sysplex (coupled) environment, it is (TP) day.also possible to allocate one or on-line the many during occursconsumption peak their partitions if CICS, or IMS as softwareeligiblesuch active. is it which on partitions the of one for consumption MSU the subtract WebSphere,canlike software for mechanism WLC).charging with compared when lower much are costs MSU (zNALC cost overall the the on reducing deployment, depending software WebSphere pricing DB2 MLCfor allows of pricing zNALC type used. software the choose carefully and consider also must We pay for IMSV10at900MSUfor software billing. At the end of migration, the have IMS V9 at 100 MSU and IMS V10 at 900 MSU 900, they then forsoftwareMSU billing. 600 at V10 forIMS pay then they 400, MSU at is V10 IMS and MSU MSU. At the beginning of the upgrade, the customer will be IMS V9 at MSU 900 and IMS v10 at 100 where multiple partitions contribute to IMS V9 costs. IMS migration is partition by partition. MSU, 1,000 at V9 IMS for pays customer the upgrade, IMS Beforethe upgrade. V10 IMS to volution of the annual MLC software bill (base of 100 in 2006) for the period 2006 -2011 2006 period the for 2006) in 100 of (base MLCsoftwarebill annual the of Evolution hey pay IMS V10 at 900 MSU for software billing. In mid-migration, IMS v9 is at 600 at is v9 IMS mid-migration, In billing. softwarefor MSU 900 at V10 IMS pay They Figure 17:IBMsoftware MLC cost evolution duringthe2006-2011period his preventsrelatedoverchargingand This other for u MU sg fo te ac wrla and workload batch the from usage MSU Thus ach new Each he IWP The They This can benefit from the mass infrastructure provided by the host (outsourcing company). h eitn ifatutr, ihr y prdn o b cagn servers. changing by or upgrading by either infrastructure, existing the for evolution of stagesseveral through go to necessary be may it sharing, server to road the On the PSLC aggregation (considering all fromthe priceplexbenefiting ~30%, servers by as costs one MLC server IBM in lowering terms while of costs, softwareMSU of incremental servers. billing). additional the number reducing without configuration, priceplex a is practicable solution only the case, of this In thebilling products). AG to Software (E.g. software server related physical certain cost, software incremental significant generate to proving ultimately beneficial, financially be not may sharing server cases, some In sharing. of level the increasing is not only to consolidate the servers in one place, but above all, to reduce the number of servers, The sharing of servers remains the most economically attractive solution. In this case, the matter a for except maintained), priceplexserver configuration consolidating the variousper servers (Before - Figurebilling 18 & After - Figure(software 19). reduction cost software no be will there costs. estatereal as well as costs, operating center centers where servers are transferred to thetarget data center. sharing. Thesavings realized server are for data without performed be can step first The • • Here, thetarget numberofdata centers can vary, for example: capitalize onexisting investments andminimizethenumberoftheirdata centers. This usually takes place in companies with multiple data centers. Consolidation withinthecompany -server sharing Finally, donotforget thecosts ofhosting… to bearallofthecosts associated withaspecificchoice. basis for technical upgrades, for financial reasons, and as a single small customer it can be hard expected. lowerbe companymight than individual for each savings hostingcompany,customers the the at between differences technical significant are there If context. this in customer “small” a al). being et upgrade, technical simultaneous (E.g. become arising process. synchronization this data their using customers other (infrastructure).facilitiescenter with hosting developments the technical for their essential) synchronize (even to necessary company be will it savings, cost software and hardware For transition isnotalways easyto manage. this and al), et Information, Billing, Support, (E.g. customer direct a as suppliers referenced important by some for consideration requires issues. still structural it interesting, sounds solution this Although Centers, asthefacility isnow withthehosting company. (E.g. company.hosting the by negotiated sharing. of be levelgenerally will acquisition the the but infrastructure, on dedicated a maintain depending typically will company variable are savings the viewpoint, hardware a From •The contracts of other software suppliers are negotiated by the hosting company, with the right •The IBM MLC software, if included in configuration eligible for PSLC pricing and using the same From asoftware viewpoint, we can note adecrease of: results inlower hardware andsoftware prices, andcan minimizesomeothercosts (E.g.people). internally bythecompany. tested then Consolidation areof 2 (or more) activities data centers; Recovery the Disaster Recovery Disaster Plan (DRP) testing service. is periodically atthe3rdperformed partysite. excellent level a and for site continuity organization party business 3rd a of involves this center; data single a to Consolidation of usebyvarious customers (economies ofscale). versions ofsoftware astheothercustomers ofoutsourcer. nergy, Facilities Management, et al) and the eradication or sometimes elimination of Data of elimination sometimes eradicationor the and Energy,al) Facilitieset Management, he extreme situation is for an individual customer to be assessed on a case-by-case a on assessed be to customer individual an for is extremesituation The e ot infcn i te os of autonomy. loss the is significant most The he more customers they have, the more complex the problems the complex more the have, they customers more The he hosting company has an important and even crucial rolein crucial even and important an has company hosting The he client will realize additional savings for operating costs operating for savings additional realize will client The owever, in the absence of server sharing, server of absence the in However, his is a simple consolidation of data of consolidation simple a is This eeoe i i nt las easy always not is it Therefore, These companies are seeking to e utmr s o longer no is customer The e atr ouin is solution latter The i aod the avoids This T ypically, it The

PAGE 51 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 52 number of engines in a single server (coupling factor)configuration.(couplingincreasingsingleserver a in engines the of number by generated degradation power CPU the consider must we hand, other the On be less thantheperfectly optimized(defacto) multi-server configuration. widespread.are but time, from differentloads partitions, previouslypeak distributed acrossif multipleespecially servers, do workloads,not occur at multiplethe same of advantage full takes solution server single This There isnoneed to priceplex (server sharing)in asingle-server environment. pricing. eligiblePSLC with associated constraints the eliminates solution single-server MIPS, a al), 53,000 et to z196 MIPS, ~32,000 to z10 (E.g. capacity server in increase continued the With to benefitfrom optimized cost viasoftware billing. As previously described, server sharing requires the presence of a PSLC eligible Sysplex structure Consolidation intheextreme before configurations), Sysplex pricing reaching theidealandoptimalconfiguration (aneligible PSLC Sysplex containing allservers). several thus Sysplex (and configurations multiple intermediate via evolve interconnectingseveral to necessary be may of it project, of type feasibility this implementing of cost the the and structures, namely constraints, technical Observing organization andoperation, generating implementation costs. implement. to trivial not generally is differences,and software hardwareand limits strictlynote operation Pleasethis “priceplex”.that or Sysplex” “billing a called commonly billing, software for Sysplex a not obtained naturally from existing Sysplex structures and partitions, we must consider is building this If pricing. eligible PSLC for required is Sysplex a configuration, multi-server a in Finally, activities require investment inequipmentandrelocation costs. fromsoftwaretechnologyAWLC,additional (E.g. invoicedividend, savings al.). et benefiting and servers) generation-faster (new performance better have to opportunity an also implementation ofa standard Sysplex andthus PSLC eligible software pricing. scale. of economies from benefitting thus used, software of amount system. this transferring to then users and and system data information target a selecting by achieved be can Convergence Banking) sector, andthey alsoexist intheworld ofindustry. merger of entities with similar activity profiles. Such cases are not uncommon in the finance (E.g. systems. of convergenceconsider to is step next the systems, informationmultiple for servers of sharing and centersdata of consolidation the After Platform convergence take longer to implement. must consider that this is a single point of failure. In this case, problem resolutions will inevitably configurations generate will failures (such configurations are architected& not to coded we fail), the to closer architectural is limits of server in that terms of number configuration of engines and power.a choosing Unlikely with as it is that associatedthese risks the consider also must We the GDPSorsimilaroption,generating continuous availability for service continuity. answer can then be a multi-server Sysplex with data and load sharing, possibly supplemented by observe that such a configuration is incompatible with the demands of continuous availability. of restarttoserver).service time continuitybackup (E.g. a to activity switch and requirements the meets it if possible only is configuration single-server of type this Finally, a in than configuration server multi-server configuration.single a in engines more be must there power, CPU equivalent e mlmnain f blig ype hs mat i trs f architecture, of terms in impacts has Sysplex billing a of implementation The herefore the MSU resource for software billing purposes will always will purposes resourcebilling softwareforThereforeMSU the i hroiain hs alw u t vldt ad ii the limit and validate to us allows phase harmonization This his convergence is made possible by the by possible made is convergence This T herefore it facilitates the facilitates it Therefore o achieve like-for-likeachieve o and Thereforemustwe However,these The All oftheactivitieshighlighted above (Figure 19)have beenimplemented, asper: (Figure 20)in cost, while incorporating theannualworkload growth activity. The software billing savings in this case also live up to expectations, representing a ~50% decrease 7,500 to 30,000MIPS). (I.E. power server Mainframe associated the in increasefour-fold a been has there period, year5 same this in Meanwhile, period. same the in ~20% only was capacity MIPS installed total the year), per (~12% period year 5 this during capacity workload increased the incorporating While • Convergence ofInformation Systems. • Infrastructure sharing. • Data center consolidation. •  •  •  These software savingswere generated from several initiatives: (Figure 18)to another(Figure 19): Starting in 2007, research conducted over several years allowed the move from one configuration Implementation case study profiles. Consolidation and extreme infrastructure sharing, to optimize resource use for different activity application workloads. the of benefit the resourcesfor machine of use better partitions, of number the in Reduction and environments technical of Development environment costs. number the reducing Systems, Information of Consolidation Number ofglobalMSUfor Billed AmountVAT/Month Number ofMLCsoftware - CPUcapacitytotal:30,000MIPS - Mainframeserverpower:totalof30,000MIPS - 30partitions - 2informationsystems(+1instand-bymode) - 1activeMainframeserverwithbackup - 1datacenter - CPUcapacitytotal:24,138MIPS - Mainframeserverpower:from568MIPSto7,500 - 53partitions - 6informationsystems - 8servers - 5datacenters Item Description (Base 100) invoices billing Figure 18:Server consolidation activity–before configuration ofApril2007 Figure 19:Server consolidation activity–after configuration ofApril2011 Figure 20:Server consolidation activity–MLC software billinganalysis Configuration April2007 Configuration April2011 April 2007 2,737 100 5 April 2011 2,282 50.35 1

PAGE 53 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 54 handling thesenew workload functions: modified System z microcode), although technically, only the MVS (z/OS) Dispatcher is capable of these processors are in fact a creative business practice (not a technological feat, it is actually just processors (CPU engines) for accommodating these new workloads. From a general point of view, T the 1980’s. of screen) (green terminals 3270 simple the than manage to difficult more now also is network software for (ISV’s) suppliers developmentAssemblerinsteadof by (HLASM). preferred commonly language, C/C++ the also but language, Legacy. as classified environments, older many replace technologies new these Indeed, financial. but technical, inherently not image. is development dinosaur its retains platform Mainframe the Nevertheless, to Mainframe Java technologies andmore recently SOA, XMLandWeb 2.0. CICS. and IMS as such traditional the complemented has middleware new The DRDA networking protocols. which generated the emergence of mixed applications, including the ability to access DB2 data via exchanges between the Mainframe and the Distributed Systems Similarly, world became more commonplace, today. competitive” “remain and technology new to up open to platform Mainframe the allowed which integration of MVS UNIX kernel APIs in 1994 (Open In recent years the Mainframe software landscape has changed considerably. It all started with the 3.2.2 Useofspecialtyengines:zIIP,zAAP,IFL o limit “clunkers” or flops, in the 2000’s IBM has been imaginative introducing specialized introducing been imaginative has IBM 2000’s the in flops, or “clunkers” limit o T PI hs hne te iuto, aig uesdd h SA network. SNA the superseded having situation, the changed has CP/IP Figure 21:z/OSDispatcher architecture schematicfor zAAPspecialtyengine he WebSphere Application Server is the most famous, opening the door the opening famous, most the WebSphereis ServerThe Application he dispersion of Mainframe customersMainframe acrossof dispersion The the his increase is mainly due to use of the Java the of use to due mainly is increase This Edition since 1998 and Unix System Services), T ) monitors (TP) Processing ransaction e btce o platform for obstacle The u the Thus 3.3 Controlling operational costs • supplier that has extensive experience andsolidreferences inthematter thuscome naturally. makers. policy to internal to obvious be even or appeal A well-managed/executed quality approach meets this double objective, but it does not necessarily costs ofinvestment. fiscal constraint. In such cases it is importantsignificant toof lowercases operating in costsparticularly and/or deferlimits, theits maximum reach may above described operation of mode The during diluting theavailable skills. occurring only deliver once approach, it can as recognized, systematic was. nor valued that a neither seemingly not is optimizations work this is Moreover, crisis. punctual operational this typically perform but technicians results, good spectacular sometimes course, Of existing resources (domore withless). our optimizing than rather budgets, operating burdenour courseforarea of growth, which with ever. essential more than (quality) approach an such makes specifically, it name to quality, development al. et schedules, constrained loss of technical team skills, personnel reductions, balancing the speed of project implementation,Over a period of time, a quality approach tends to lose its importance, for various reasons, such as “The fault,dearBrutus,isnot inourstars,Butourselves…” -WilliamShakespeare (JuliusCaesar). 3.3.1 Optimization,performance &technical/application component quality •T latest z196server: clearly visible. and is attractive its for interestingoptions new two announced has and momentum newapplications its continues IBM That’swhy of these impact financial the Nevertheless, far from negligible, which explains the adoption of this solution by a large number of businesses. to zIIP, but such usage remains low. agreement from IBM. CA, Syncsort, and others have exploited this function, offloading CPU usage prior with requests),function, zIIP exploitthis can DRDA(ISV’s), vendorssoftwarewho other to open ~50% also are but (processing workloads DB2 for attractive also are processors zIIP The is now becoming animportant metricto consider duringassociated tuningactivities. CPU of ~40-70% consuming usage, depending workloads, on Javathe type for of application. intended primarily are processors zAAP The also which engine), specialty a limits theoverall General Purpose(CP)CPUusageimpactfor othersystemon software partitions. cost no for runs software (arguably cost MLC from exempt However, the most important benefit concerns the software billing invoice. accommodate such new workload applications (E.g. DB2 DRDA, Web 2.0, et al.) on the Mainframe. approximately a factor of 4 when compared whereto processors, standardspecialty new processors, these it Without was not financially feasible to by ~30%,while exploiting thezIIPspecialtyenginefor optimizedCPUusage. will be handled by the native XML APIs now included in the MVS nucleus, improving performance On the software side, DB2 Version 10 offers new perspectives via XML databases. Indeed, requests evr s eShr, otd n h sm priin s h lgc dtbs sses and systems database legacy the as partition same the associated T in hosted WebSphere, is server unrivalledmostadvancedservice,and of performance, quality applicationespeciallythe when an provides still which environment, Mainframe the within function server application this of CICS. and processingIMS monitors, (TP) significantly conventional the can with WebSphere by generated that SCRT, with associated deduction IWP and AWLC metrics, applications. eligibleMLC these for pricing billing reducesoftware MLC new are there Commercially, specialty engine,inconjunction withthezAAPengine,asandwhenrequired. .g. Java, DB2, et al) on the zIIP the on al) et DB2, Java, (E.g. workloads all consolidate to possible now is it echnically, he task is further complicatedtaskfurther The is mergerthe with consolidationor severalof further entities, owever, we often prefer to make investments (hardware and software) associated software) and (hardware investments make to prefer often we However, ransaction processing (TP)monitors. his situation is quite paradoxical, because the steady decline in decline steady the because paradoxical, quite is situation This However, the financial gain associated with DB2 workloads is The ratio associated with Java/DB2, easily identifiable, his is a breakthrough that should allow a wider use wider a allowbreakthroughshould a that is This T otal Cost of Acquisition (T Acquisition of Cost otal he idea to get help from a specialized a from help get to idea The his new mechanism allows MSU allows mechanism new This These processors are CA) was reduced by reduced was CA) T ransaction

PAGE 55 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 56 Technical considerations, before project initiation: project before considerations, Technical initiation: project before considerations, contractual & Financial The primarykey success factors (pros) andpitfalls to avoid (cons) customer to develop andmaintain attheirown expense. Sotherethe isreal value forfor thecustomer. expensive too be would generallythat analysis/mining) providers data (E.g. specialist tools and these experience skill, Additionally, have cost. delivery the limiting of guarantee a all providerspecialist forthe chosen. approach. An original or innovative business model can help, such as result based compensation Therefore we must extensively explain, argue and convince for the need and cost effectiveness of the (system &application), whichthey usually donotneed,andwhosecost must beconsidered. teams technical for workload additional an generates cost, intrinsic its excluding approach, an profitability, such growthusefulnessand conventional a Indeed, its of solution. compared that to The first difficulty encountered in the implementation of such a quality approach is to demonstrate Structuring theproject have every chance ofbeingrejected because oftheirburden andassociated risk. code). application in is committed wherever possible to offer only easy to implement optimizations (I.E. minor changes Optimizations for the application base code are usually in the majority and the specialist provider the and al.) et DFHSM, WebSphere, CICS, application IMS, portfolio. DB2, as such components major and System The specialist provider offers optimization for both the technical foundation base (z/OS Operating the necessary tools andimplementation assistance. serviceproposedproviderthe scheduledthe so, If do tooptimizations. assistscustomer the with of implementation or any other yardstick. It is then the responsibility of the customer to acceptrejector implementproposals,these expected involved, the risk on the based benefit, the complexity optimizations. of number proposingcustomer,a the with By agreement between the two parties, the provider will conduct a several month review process commercial offer. improvements. for potential the has customer it the provider, that specialist clear is the by situ in out carried analyzes and measurements of set first a After Principle oftheapproach provided are correct. figuresthe guarantee somehow must IBM matter, this in jury and judge the projected.Although In conclusion on this point, it is essential for IBM to confirm and validate the financial simulations • • • • he financial compensation package and method for remunerating the specialist provider specialist the remunerating for method •T and package compensation financial he For customers with a standard IBM therefore be appropriate to align the specialist service with an amount. Such a MSU reduction may limit or cancel the financial benefit of optimization. It may provider hasincludedthemintheir delivery package,for noextra cost. for provider specialist unpleasant surprises that either these the tools exist within the customer, or that with the specialist inventory require. they an tools diagnostic prepareand measurement any also must we terms, practical In savings and savings thatitexpects to make for their customer, settingtheir prices the accordingly. of the evaluation in provision. budget associated both disappointments, significant to lead then can for example by dividing the bill by the number of VWLC MSU consumed. Such a linearization MSU, average the on relyprimarily not do So cheapest). (the marginal be will savings cost MSU the usage, MSU higher for So metric). cost use/lowest highest and costlowest use/highest the between ratio 1:10 a (E.g. cost MSU to related scale sliding a has pricing MLC evaluationprocess.expectedsavings the toclose attentionPayPlease rememberIBM that facto, ipso thus and removing allworthwhile substance.scope, Goodnegotiationshouldoccur priorto anyagreement. its limiting clauses by accompanied be may performance for pay of principle the as provider, specialist the by offered contracts the consider Carefully must be kept simple andlegible, beingcontemporaneous with current financial processes. hey know that complex optimizations are implementation prohibitive and prohibitive implementation are optimizations complex that know They his can be both a guarantee of their commitment, and aboveand commitment, their guaranteeof a both be can This his is essential, because the specialist provider anticipates provider specialist the because essential, is This ESSO contract; the MLC cost cannot fall below a minimum he customer then has the freedomto the has then customer The his check is necessary to avoid any avoid to necessary is check This hey are then able to present a present to able then are They ESSO contract renewal process. vrl, yial sc otmzto dlvr te rmsd eut. n ey ae ae, results cases, rare the of competence very the also but measurement, of In importance the hence expectations,below results.are promised the delivers optimization such typically Overall, The proposed optimizationscan beclassified into three broad categories: Optimization methods–technical review savings the because invoice, software associated found mightbeabsorbed,ifwe don’tmake the required the end-to-end changes. with materialize not will the realizedOtherwise reduction). gains MSU (I.E. point that to realized gains associated the per as capping, weprogressAs through deploymentthe essentialis it optimizations, toof lower levelthe soft- of active participation of all Production operational decision-makers, from the ensure all to of essential the is technical it teams.benefits, optimization timelysafeguard to previouslymentioned, As to currently space” give “breathing can they experienced Production constraints. Moreover, profitability. deployment project impact directly benefits financial optimization resulting the because essential,& is optimizations of implementation rapid The Implementation Given theseobservations, we shouldpayattention to: choosing the not thus and wrong IBMMLC billingmethod. optimizations, of proposed impact financial the verifying essential, preciseSimilarly,a MLCunderstanding IBM principles of billing fromabsolutelycustomer is the will bemore beneficial,especially inthe long term. approach the optimization, of area the in competence and maturity demonstrate team technical customer the if short, In provider’steam. consultancyspecialist the with pragmaticallyinteract have an accurate reports. analysiscomplete and reliable if with they associated view a service, Production the customer their to related for issues performance of more view substantial the all is benefit cost Generally, initiation Project • their provider,fromspecialist the •Twhich for ourselves,and found have not could we hose or negligence of because maybe or time, to due implement •Tnot do we but know, we hose • • • • • spending apenny. projects, without and migration/optimization easily quite performed for be can migration/optimization price maybe fact, in high whereas a charge companies some that discover topics. these for projects longer-term manage eventually optimizations. specific All that we are told informally in terms of good practice, experience atothercustomer’swe andassociated expertise, addsrealfollow, value. but they are not necessarily supposed complexity ofimplementation. is then to quantify the technical and financial gains that we could obtain, and to eradicate the laxity. Unfortunately, these are the most numerous. easily controlled andreused later. moreprovider. be specialist preferredwill the tools used by those Procedurestools to and Indicate as much as possible the use of customer installed measurement tools, as opposed Agree onamethod for measuringbenefits,andmaintain control onit. handling withtheirown resources. capable of not is customer the that activities on mainly work infinitely to more provider. beneficial is the specialist by It assistance. without undertaken future, the in and currently both possible,of investigations optimizations scope the Confine responsibility to monitor thischangeandgeneralize accordingly. processings. other many to applicable equally be can detected and implemeted on processings impacting the peak load, hence onlybut Fordistinctions.simpleexample,optimization, some a make to the have may they but software billing, bill. software the reduce will that pick) (cherry optimizations priority on focus will provider specialist the that forget not Do inherent inthisapproach, actively participatinginitsimplementation. Management, workload extra the preparefor can Productionthey that so al.), Development, et Operations, Management, Systems Analysis, Systems (I.E. units customer internal those all service, the of start the before and earliest, the at engage to important also is It wvr tee vne ms b epoe, loig h cin to client the allowing explored, be must avenues these However, qually the customer requires competent technical staff able to able staff technical requirescompetent customer the Equally his is consistent in most cases with customer’s goals, customer’s with cases most in consistent is This The real value of the specialist provider eeoe i bcms customer a becomes it Therefore, his is often an opportunity to opportunity an often is This T d s, ulns the outlines so, do o

PAGE 57 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 58 • •AddingMainframe environments (E.g.T partitions, Mainframe workload hosting techniques are alsochanging: In parallel to changes in servers, environments (E.g. •Increase total installed MIPScapacity (~10%peryear). •Increased capacity ofinstalled Mainframe servers (+400%for someusers). changed thecharacteristics ofMainframe servers inrecent years: significantlyresourcehas IT sharing infrastructureand consolidation, consolidationcenter Data 3.3.2 CapacityManagement,tools &methods performance, whichsadly isseemingly opposedto thecurrent dominanttrend. Mainframe to related expertise of field unique the reclaim a and relegitimize is rehabilitate, provides, to it opportunity benefits the beyond here, described approach quality the event, any In themselves. basis to deliver such benefits, so as to correct other abuses that they have been unable to correct If the customer fails to be proactive, they will have to rely for such a specialist service on a periodic determination. Thismayrequire resources (includinghuman)thatthey must acquire. and rigor with them applying imparted, practices good the sustain and integrate to be reoccur.then will not do they that ensure to steps take also must they both deficiencies, technical organizationaland commercial. endemic If broader the highlight customer is often to optimizations learn proposed from The this experience, correcting dysfunctions, Durability oftheapproach year. activities. upgrade capacity CPU routine and usual the defer to mainly but conditions, initial and growth particular the on dependent invoice),software MLC thereforethe (and reduction usage MSU an was there approach, this followed that all For Benefits service (socalled routine errors). analysis rushed a to linked typicallyare rarelyfortunatelyexperienced, disappointments, These customer’s technical teams, which allow for a debate on equal footing with the specialist provider. MLC software invoice billing level, but also types of contracts implemented (E.g. IBM the onlyimpacted not associatedevolutionsharing evolutionswith server Mainframe These This process can berealized asfollows (Figure 22): Management for Mainframe operations thatintegrates alltheproblems andrequirements. T such asthe“CapacityPlanner”and“Software Cost Planner”. financial impacts for Mainframe datacenters generated the development of emerging professions techniques. optimization costsoftware and hardware for search constant a is developments these from resulting objective major The OIO, etal.) o perpetuate this search for cost control and infrastructure, we require a process of Capacity of process a require we infrastructure, and control cost for search this perpetuate o Removal of Mainframe environments (E.g. he rapid growth and associated hardware and softwarehardware and associated and growth rapid The est, Production, etal.)whensharingservers. T est, Production, et al.) when consolidating servers. Figure 22:IBMMainframe capacity managementprocess T est, Production, et al.) and their associated his benefit is generally enjoyed for one for enjoyed generally is benefit This he challenge for the customer the for challenge The ELA/SRA, ESSO, T • • • this activitywe can (figure 23): of end the calendar),at their and per customerrequiresas wheneverthe so (or do yearto the of takesupdateactivity Planning Capacity This place onceyear, 1sttypicallya the of end the at updates, periodic per requirements resource and definedserver configurations tobe associated with capacity 3-year requirements. as of establish, projections quarterly to situation, current consolidated the to subsequently corresponding are elements these of All for entitiesandserve asabenchmarkwhenchangingworkload behavior(pros &cons). users. It is even quite difficult to retrieve indicators of activity targets that may yet justify the need end- our of plan business the toaccess havelinks not or do we if disappointing level,is this it At for considered be to rate growth the as future years. well as needs, these the justifying documenting events, requirements, key resourceupcoming annual gathers function management capacity The is aresponsibility oftheseotherentities(technical teams). doubt, in any case, capacity management is not a substitute for performance management, which operate the Mainframe platform, who are obliged to provide services for their end-users. Without The capacity management process is a unifying element for the various entities that administer and and administer other otherfunctions processingmanage onMainframe servers. that to address entities associated or teams operational and not intended the of responsibility is the management to more it pertain capacity that issues to note; limited Please techniques. is configuration here server presentedMainframe process configuration the of role includes The generally and servers management performance management, Mainframe capacity management. problem management, of change management, management Comprehensive Capacity Management–Role &Scope an infrastructure consolidation project. o explain this process, we have provided an example of Capacity Management implementedfor Management Capacity exampleprovidedprocess,of havean this explainwe o upgrades andIBMMLC software billing. Develop the capital budget and operating costs for the following year, both in terms of hardware Define server configurations and scenarios. Observe the 3-year trends and projections (accompanied by supporting and working hypotheses). Figure 23:IBMMainframe capacity cost planningprocess Half

PAGE 59 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 60 • The first step is to develop adocumentidentifying requirements: CPU (MIPS)CapacityPlanning Capacity Management:Process Overview and hardware center data Mainframe of software costs. control optimal for allowing process, management capacity the processes,of (E.g. al.) procedures,scope et the monitoring, outlined tools,We have situation mayaffect investment budgetsorbusiness operations for thecurrent year. requirements.down) or (up change associatedaccommodate the to plan capacity the Production, et al.) in case of drift, configuration (redistribution of resources) or trigger a refresh of the review, it may be necessary to modify a Mainframe partition or an environment (E.g. environment an or partition Mainframe a modify to necessary be may it review, the establish a regular monitoring of Mainframe behavior, reconciling estimate projections. Following Because capacity management forecasts are based on best endeavors estimates, it is essential to also used regularlyis aspartofsoftware contractprojections renegotiation andsoftware acquisitionactivities. year 3 with associated process management capacity the from Information changes. unidentified configuration changes,hardware or evolution andchanges to theIBMMLC software exceptional bill. to due required, if year, and the as during replayed be can hypothesis working associated with processforecasting This example: for 25), Figure & 24 (Figure years 3 next the for projections quarterly establish and year this of From these requirements, it is possible to update the MIPS capacity requirements for the 2nd half •For servers andassociated partitions. By environment (E.g.T - Thememoryrequirement, perpartition. -  - Thecurrent MIPS usage. - Requirements andtheirdeadlinesfor thefollowing year (andbeyond ifpossible). - Thegrowth rate (E.g.annual). - Changeassumptions duringthenext year asaminimum. of aspecific requirement, thegrowth rate willbeapplied. MIPS Specific requirement possible). absence beyond if the the (and forthcoming for In year Figure 24:Sample 3year IBMMainframe server MIPScapacity forecast spreadsheet Figure 25:Sample 3year IBMMainframe server MIPScapacity forecast histogram est, Production, etal.). i cn eeae Mainframe generate can This his latter This T est, ne gi, h zC to i vr ueu ad ruby adtr t ahee hs different these achieve to mandatory arguably and useful very configurations and comparisons. is tool zPCR the again, Once differencestechnology (E.g. z10,z196,etal.). the also and server) Mainframe single a for is example this (I.E. center taking into account the different models of Mainframe servers, the number of servers in the data configurationsof simulations possible,performnecessary,evento is it stage, this At 2013. early in technology server Mainframe change to requirement a highlighted have we example, this In requirements and space) MIPS (free MIPS the available identify anybreak the points. with highlight compared to process, then Planning Capacity is the capacity in forecasted actual activity, zPCR the Following Capacity evolves according to theseparameters. capacity.serverMainframe on engines specialty logicaland partitions, engines of number the of tool. zPCR IBM calculatedthe with is capacity server Mainframe the 28), Figure & 27 (Figurepreviously presented scenario the In evolutionof the define and to size) servers. capacity MIPS model server (specific points breaking any servers. Mainframe forecastedidentify requirements,the the meets 28) Figure configure & 27 (Figure capacity and server Mainframe size to is step second The Server configuration development A similartable isalsodeveloped for Mainframe memoryrequirements. the PR/SM configuration parameters for the different Mainframe servers hosting these partitions. developto necessaryare 26) (Figure partition per MIPS requirementsdetailedidentifying These Figure 26:Sample 3year IBMMainfram server/partition MIPScapacity forecast spreadsheet Figure 28:Sample 3year IBMMainframe z196server MIPScapacity forecast spreadsheet Figure 27:Sample 3year IBMMainframe z10server MIPScapacity forecast spreadsheet his tool has the advantage of taking into account the impact accountthe into taking of advantage the has tool This i se cek ta the that checks step This

PAGE 61 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 62 will allow to: SLA), or AgreementsLevel Service (E.g. indicators service with comparison and analysis this of changes. these of origin the understand to analyzesdetailed morerequirement for a The findings associated with current period will identify possible changes (up or down); generating a and forecast, the from platform, Mainframe minimum ofhistorical information. the of behavior current review to able be must assumptions, sometimes fluctuating, sustaining the process and delivering benefits. The MIPS Capacity Planning process is essential to establish a monitoring process, developed on CPU (MIPS)apacityPlanningFollow-On taken into account chargeable inthebudgetevaluation activity. be to software the of version one table.formalizedthis not SVC),in is IBM tonegotiatewith migration(time the during only for allowing activity, SVC an of impact The The MSU Capacity Planning process only formalizes the billing period of two versions of a product. for anSVC (E.g.1year software holiday,to migrate software form oneversion to another). Of course, asoftware upgrade generally involves achangeinpricing and negotiatingwithIBM (E.g. the server configuration PR/SM set-up) made available to partitions and Mainframe environments Previously we used the MIPS metric to detail the amount of CPU resource (resources required by CPU (MSU)apacityPlanning and configurations server forecasting when account into taken be requirements, to enactthescenario, integrating thesedecisionsinto theinvestment budget. will elements these All •IBMESSO contract management(orimplementation) assistance. •Forecasting theestimated budgetfor next year. •MSUmanagementtechniques (E.g.soft-capping, etal.) •MLC software billinglevels, byproduct (E.g.z/OS,CICS,DB2,etal.). • This table (Figure 29)willtherefore highlightthefollowing: processes, contractual commitments, etal.). cost software optimized implementation, soft-capping (E.g. billing software for policy company the (especially) and MIPS Planning Capacity the account into taking by calculated is growth (%) determining an annual percentage (%) increase applied to the base (1st Quarter). This table is categorized by product on a Quarterly (3 Monthly) basis. From these elements, we perform the MSU Capacity Planning processing for the next three years. dividend, technology z9/z10 (E.g. usage MSU impact will that •Hchanges technology ardware •The current level ofdifferent IBMMLC billingsoftware (starting point). •Soft-capping implementation. • • • The MSUCapacityPlanningprocess outlinedhasbeendeveloped considering: the implementation ornotofsoft-capping. tool. SCRT IBM the 4 (Rolling R4HA the of notion the consider we Software changes (product additions,removals orversion change). z196 AWLC, etal.). Forecasts ofthesoftware products usedandtheirassociated version numbers. Forecasts for theevolution ofactivityontheseMSUusagevalues. IBM MLC eligible software usedondifferent active zSeriesservers andassociated partitions. • T usage withno increase in activity), to return to asituation classified asnormal. Make corrections, for example at a Mainframe environment level (E.g. abnormally high CPU est, Production, et al.), while MSU are used for IBM MLC software billing purposes. So here Figure 29:Sample 3year MSUcapacity requirement, byMLC software product spreadsheet he calculation of this value can be generated in different ways accordingdifferentto ways in generated be can value this of calculation The our Average) MSU usage, calculatedmonthlyforusage, Average) MSU Hour The values are calculated by The percentage This process his will be will This he result The the amountofinstalled MIPS,only theusage level. ¼ hour (15 minute intervals), it’s not an instantaneous peak. Finally, this graph does not represent reached in the month during the on-line (TP) day consumptionperiod. Sinceof levelthe maximum measurements the areand levelmade (CPC) every server a at usage MIPS displays graph This (E.g. environments and servers Mainframe for usage MIPS peak basis, monthly a example,on this In (E.g. time over annual capacity plan. checking consistency for allows tendencyto overstate and requirements, informationSuch al.). alwayset is the updating when useful trends displays information Historical importantly, to control andsustain investment. most and time right the resourceat requirementsadjusting behavior), in abnormalities certain correcting and (detecting Plan Capacity the to change any control to important is analysis This between 10:00-12:00duringtheon-lineT • • T g atcpto o an of anticipation (E.g. upgrade, anactualupgrade ordeferral, etal.). roadmaps developments infrastructure and investments Modify or decreasing (E.g.requirements over-estimate, decreased activity,etal.). change usage and associated forecasts. Modifications may be increasing (new requirement) permanently will hypotheses new because Planning Capacity of exceptionalupdates Make s, rdcin e a. i maue. n h sml gah eo, h pa i positioned is peak the below, graph sample the In measured. is al.) et Production, est, Figure 30:Sample several year CPCpeakMIPSusagefor theon-line(TP)workload graph ransaction Processing (TP)period:

PAGE 63 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 64 the on-line(TP)peaksevolves upwards over time. increased of phase significant 2011. January in exceptionalactivity an of and a peak 2010 June in activity seen has partition PF2 the 31), (Figure above example this In e aibe rnlrt o ifrain los s o eiy h cnitny f S Capacity MSU of levels andassociated anticipated budgetary requirement consistency (upordown). the verify soft-capping necessary) to if (implementing differences,adjusting behavioral us identifying Planning, allows information of granularity variable The and partitions,establishing billingprofiles andthepeakdayduringmonth. to possible is (CPC) servers Mainframe it of usage the al.), per as billing, software et MLC of levels actual SMF, the monitor SCRT, (E.g. sources information various from data collecting By CPU (MSU)CapacityPlanningFollow-On a complete shutdown duringJuly 2011. In this example above (Figure 32), the PF10 partition has experienced a usage decline in 2010 and Figure 33:Sample daily MSUR4HApeakfor SCRTbillingmonth(start of2ndday-end1st day) graph Figure 32:Sample several year PF10partitionpeakMIPSusage for theon-line(TP)workload graph Figure 31:Sample several year PF2partitionpeakMIPSusagefor theon-line(TP)workload graph he trend line based on trendbased The line Now for anexample where theproduct version numberchanges: planning: capacity and billing quarterly average the billing, monthly forlevels the show below graphs the producedmonth, everystatisticsbeing SCRT With months. 2 between strongfluctuations to due reviewprocessmonth) quarterly (3 a per evaluatedas producedand was plan capacity MSU The billing amount,we only needto actuponthe12:00-13:00(TP)timeperiod. day. the during once only reached is maximum This previous 4hours oftheon-line(TP)workload (I.E.08:00-12:00). period. time 12:00-13:00 the in is (R4HA) amount MSU billing maximum The In thisexample above (Figure 34)we can deduce thefollowing observations: it willbenecessary to actsolely onthepeakday,asfar aspossible. month. the in once Other peaks onlyare reachedusually 2,000-2,200 MSU. was If the customer MSU) wishes to reduce(>2,700 their softwaremaximum billing, this that and month, the of day In this example above (Figure 33) we can see that the maximum R4HA peak is reached on the 1st Figure 36:Sample MQV6&V7(SVC) usage,asperSCRTcalculations, graph Figure 35:Sample z/OS&DB2MSUusage,asperSCRTcalculations, graph Figure 34:Sample MSUR4HAfor peakSCRTbillingdaygraph herefore to control and reduce the MSU the reduce and control to Therefore his is linked to the linkedto is This

PAGE 65 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 66 e otiuin e evrnet (E.g. environment per contribution The processes. each for share the display chargeback for perhaps software, MLC and key define to used values attachedbe relevantcan information Such environment. the being derive partitions to easy With is partition. it each then environment, of an contribution to the and billing z/OS of level the No alert,billinglevelconsistentwithCapacityPlanning Alerts andWarning Points Figure 37:Sample SCRTMSU&associated workload share (%)analysis table Total PF9 PF7 PF6 PF5 PF4 PF3 PF2 PF1 PF z/OS MSUsperPlatform T s, rdcin e a. i dtrie (iue 7 from 37) (Figure determined is al.) et Production, est, Consumed Consumed 2 716 2 716 1 189 224 956 192 17 44 15 79 CP MSU 2 743 Contribution in% 100% 35% 44% 8% 7% 1% 2% 1% 3% -0,98% Gap 4 CONL is notahistory oftheMainframe, thisisahistory ofcomputing. document, this in articulated have we What exploresolutions. and questionssame same the the discussed in this document, will be replicated for Distributed System environments. ahead, we can make a bet; in ten stepyears, in fiveone years,often in three document seniority, years, the costof controldecades our steps thatits we with to specialty, returnMainframe the review we we if If introduction, tomorrow. for templates create helps also ever-lasting. it and importantly, competitiveMore more them make helps control cost platform Mainframe Better still improvements, for leeway some avenues always to explore andwork to deliver. is there that apparent readily is it approach, this supplierslicensing(ISV’s),with models are constantly evolving, reinvented.being or changing, In of many elements, including cascading and interlinked costs. Still others, such as negotiating new Others, such as the consolidation of physical machines, are the first step of an extensive overhaul weakness.areasknowledge of analyzingresourceforidentify and toexpertise allowingus used, Some activities, for example, workload optimization, require us to have already reached a level of where construction takes timeandmust beadapted for thecontext ofeachindividualuser. comprehensivesolution, a of part a only levers,each economic many aretherecontrol, cost for angle. one than it is necessary to unravel more each aspect, one after another, and sometimes all together. from Ultimately, consideration demands and Because the cost is itself the product ways of a complex cost set of elements which aggregate and interact, that many understood in have addressed will reader is the control document, copious this of end the reached Having USION They will ask

PAGE 67 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER 5

PAGE 68 Enterprise System Connection ESCON interoperability standard for DB2. Architecture Distributed RelationalDatabase DRDA storage ortiering. hardly accessed tapes. Itiscalled hierarchical ratio. For example, slow discswillbeplaced on relevant supportinterms ofcost/performances simplifying data positioningonthemost storage managementtool whichaimsat Data Facility DFHSM IBM’s DBMSoften usedonzSeriesmachines. DB2 operations onzSeriesmachines. A co-processor dedicated to cryptographic Crypto Express 3 paying only whenthey activate them. resources (processors, memory,storage, etc.), spontaneously activate dormantphysical contract modelinwhichtheusercan Capacity-on-demand CoD while savingresources. large number oftransactions simultaneously IBM’s transactional monitor able to managea Customer Information Control System CICS MLC license for z196machinesonly. Advanced W A generations andoperating environments. They allow thedifferentiation ofhardware hardware features required to runanOS. Architecture Level Set ALS AND A GLOSSARY OFTECHNIC APPENDIX FICON). transportation onMainframe (replaced by architecture andprotocols usedfor I/O The asterix (*)symbolindicates terms relating to licenses andcontracts. WL C* orkload License Charge CRONYMS USEDINTHIDOCUMENT Hierarchical Storage Manager AL TERMS Fiber Channelfor I/OlinkonzArchitecture. Fiber Connectivity FICON Mainframes. IBM’s “payfor whatyou use”methodfor entry Entry W EWL commitment. generous discounts inexchange ofa3-year all-inclusive Software contract by IBM. IBM offer Enterprise Services andSoftware Option ESSO* synchronization inaSysplex environment. for safe data sharing andz/OSpartitions partition which only runscode neededmostly Internal CouplingFacility ICF finer andmore dynamicmanagement. evolution ofPAV (seePAV below) whichallows HyperPAV and manageCoDresources. dynamically reconfigure LPARs, and to activate of logical partitionsand profiles, to HMC allows theconfiguration andmanagement SMP ofthedifferent IBMservers families. The based onLinuxfor partitionedsystems and Hardware ManagementConsole HMC that mayrundifferent OSfamilies. T hypervisor. Thegoalisto make thesameas between partitionshosted bythesame high-performance communication technology HiperSockets severe disaster. grant business continuity even inthecase ofa automated recovery to abackupsite inorder to Parallel Sysplex extension whichallows Geographically Dispersed Parallel Sysplex Geoplex/GDPS pricing. pricing), for products noteligible to usagebased fixed IBMpricing(vs. vWLC: usagebased Flat W FWL CP/IP connections in memory between LPARs C* C* orkload License Charge orkload License Charge International Product License Agreement IPLA* today. management system since thenandisstill used It hasbecome acomprehensive transaction hierarchical database developed in1966byIBM. Information ManagementSystem IMS pricing isattractive. processor dedicated to Linux for System z. Its Integrated Facility for Linux IFL based onusage. absorb workload peaks.Itisquarterly invoiced deactivation ofprocessors andmemoryunitsto pricing modelwhichallows theactivation and On OffCapacityDemand OOCOD* Midrange W MWL a system. statistical value whichindicates thereliability of Mean TimeBetween Failure MTBF depending ontherecorded MSUconsumption. large family oflicenses invoiced monthly Monthly License Charge ML systems. is suchthatthey can beconsidered asdistinct isolation between LPARs onthesamemachine Sysplex can becomposed ofseveral LPARs. The hypervisor PR/SM.ASysplex oraParallel server. OnSystem z,they are managedbythe makes amachinesubsetappearlike adistinct Logical Partition LPAR mode FCP) whichallows access to SANs. VM (hosts hundreds ofservers) orFICON(FCP processors inLPAR (z9,Z10),innative orwithz/ on Mainframes along withz/OS.Itcan useIFL for series(Architecture 64-bit)whichworks Linux OSfor S/390(Architecture 31-bit)orLinux Linux for System z sharing). code access, redistribution rights,modification model andwhichhasaFree license (source its principles), developed onacooperation Operating System inspired byUNIX(itadopted Linux maintenance). an O products. MLC license for VSEandtwelve middleware C* T C* C typelicense (purchase and orkload License Charge

English. environment (E.g.T Platform isaFrench term, whichequates to (production, pre-production, development, Etc.). outsourcer, orto different typesofactivity associated withdifferent customers for an server usually hosts several platforms functioning Information System. AMainframe stands for allthepartitionsdedicated to awell- Platform of several I/Ocommands. technology whichallows concurrent processing Parallel Access Volume PAV including purchase andsupport. One timecharge: basiclicensing model OTC System z10. interface card with410GbEthernetportsfor OSA Express 3 service will have to beavailable again. disaster orthedisaster registration, theuser 2-hour RT registration. Itistypically atime duration: a recovery length after adisaster ora disaster Recovery Time Objective RTO will berecovered. recorded until15minutes before thedisaster duration: a15-minute RPOmeansthatthedata in case ofanincident.Itistypically atime objective for themaximum loss ofdata tolerated Recovery Point Objective RPO instantaneous MSU measurements). during thelast four hours (average ofthe48last partition’s oraplatform’s resource usage unit ofmeasurement byIBMbasedona Rolling 4HAverage R4HA one. whether they includeseveral machinesorjust license for Sysplex configurations (cluster), Parallel Sysplex License Charge PSL tasks to Power7 blades. main processing controller anddelegates some In thisconfiguration, theMainframe is still the blades willwork asaMainframe subsystem. Center Extension linked to theMainframe, these machines. Once installed on a zEnterprise Blade processors byIBMwhichruntheirUNIX server “blades”equippedwithPower7 RISC Power7 Blade C* * O meansthattwo hours after the est, Production, DR,etal.)in

PAGE 69 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 70 system. S-A peripherals (hard drives most ofthetime)to a computer interface to linkmass storage Serial Advanced Technology Attachment S throughput andless physical constraints. interface. Itreplaces parallel SCSIwithbetter (very muchusedfor storage) onaserial computer interface whichsetsSCSIcommands Serial Attached SCSI SA AEWLC, WLC orEWLC). machines withsomeMLC licenses (AWLC, MLC license only for WebSphere MQon Select Application License Charge SAL availability purposes. only onesystem imagefor performance and/or clustered machines (upto 32)working with is theapplication ofthismodelto several system inalogical unit.Parallel Sysplex of several processors working asaunique a System Complex orSysplex istheassociation Sysplex &Parallel Sysplex (zO maintenance for IBMIPLASoftware products Subscription andupport S& options. Most oftheMLC licenses have subcapacity of aprocessor andnotonawhole processor. possibility to invoice aSoftware product onpart Sub Capacity* memory chips. longer onrotating magneticsurfaces, buton hard drive withoutmoving parts,storage isno Solid tate Drive SS calculated ona4-houraverage (R4HA). MSU consumption (called DefinedCapacity) Software usage. EachLPAR hasamaximum A feature whichallows billingcontrol based on Soft-Capping* WebSphere, etc.). blocks (security,development, Linux,SAP, Software and supportaround bigfunctional A packageoffering thatincludesHardware, Solution Edition* (expected performance, practical details, Etc.). also settles theattributes ofaservice contract level between aprovider andacustomer. SLA A contract thatdetermines therequired service Service Level Agreement SLA ATA D T C* C).

T A replaced A T A. functions. modified theirSoftware products to exploit their Since zAAPs were released, several ISV’s have are linked, preventing software cost increases. number ofMSUstheplatform to which they processing. Theseprocessors donotmodifythe designed to undertake Java and XML specialized processor withaspecificmicrocode, W WL OS, UNIX,Windows, etc.). execution between heterogeneous platforms (z/ particular information exchange andtransaction which usesqueues.Thesetools allows in service between applications inmessage mode The new namefor MQSeries,acommunication W Variable W VWL products for z/OS. billing suchasDB2,IMSandCICSorTivoli units, andusedfor someSoftware product MSUs, thenumberofengines,orotherbase unit ofvalue byIBM,relying onthenumberof Value Unit VU* license for z/OSsupportonly. System zLifecycle Charge SzL such asSAP, Siebel, PeopleSoft andothers. which runsomeapplications qualifiedbyIBM, z/OS license withdiscounted price for LPARs System zNew Application License Charge zNAL Software products. extended to othertasks from IBMandnon-IBM some tasks linked to DB2execution. zIIPs were initially designed to relieve classic processors of specialized processor withaspecificmicrocode System z Integrated Information Processor zIIP (zSeries 800,z890,z9BC&z10BC). monthly license for smallMainframe machines zSeries EntryLicense Charge zEL 64-bit compatible. internal architecture ofrecent zSeriesservers, z/Architecture System zApplication Assist Processor zAAP Architecture 64-bitmode. for machinesrunningz/OSorz/TPFin monthly billingmodelbasedonusage orkload License Charge ebSphere MQ C* C* C* C* C* orkload License Charge

system orcredit card transactions. example, AirlineorT with avery hightransactional throughput. For for Mainframes, usedmostly for application transactional only operation system (real time) generations. DOS/360 whichequippedthefirst machine another Mainframe OSbyIBM,successor of z/VSE hypervisors developed byIBMinthe1960’s. most recent generation ofMainframe z/VM z/TPF (Transaction Processing Facility) programs execution ontheplatform. Unix System Services (USS) whichallow UNIX MVS (Multiple VirtualStorage) capacities with OS isthesuccessor ofOS/390andcombines roadmap whichwasborninthe1960’s. z/ 2000 andsuccessor ofanOperating System operating system for Mainframes launchedin z/OS rain companies booking

PAGE 71 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 72 T able figures Figure 37:Sample SCRTMSU&associated workload share (%)analysis table Figure 36:Sample MQV6&V7(SVC) usage,asperSCRTcalculations, graph Figure 35:Sample z/OS&DB2MSUusage, as perSCRTcalculations, graph Figure 34:Sample MSUR4HAfor peakSCRTbillingdaygraph Figure 33:Sample daily MSUR4 HA peakfor SCRTbillingmonth(start of2ndday-end1st day)graph Figure 32:Sample several year PF10partitionpeakMIPSusagefor theon-line(TP)workload graph Figure 31:Sample several year PF2partitionpeakMIPSusagefor theon-line(TP)workload graph Figure 30:Sample several year CPCpeakMIPSusagefor theon-line(TP)workload graph Figure 29:Sample 3year MSUcapacity requirement, byMLC software product spreadsheet Figure 28:Sample 3year IBMMainframe z196server MIPScapacity forecast spreadsheet Figure 27:Sample 3year IBMMainframe z10server MIPScapacity forecast spreadsheet Figure 26:Sample 3year IBMMainframe server/partition MIPS capacity forecast spreadsheet Figure 25:Sample 3year IBMMainframe server MIPScapacity forecast histogram Figure 24:Sample 3year IBMMainframe server MIPScapacity forecast spreadsheet Figure 23:IBM Mainframe capacity cost planningprocess Figure 22:IBM Mainframe capacity managementprocess Figure 21:z/OS Dispatcher architecture schematicfor zAAPspecialtyengine Figure 20:Server consolidation activity–MLC software billinganalysis Figure 19:Server consolidation activity–after configuration ofApril2011 Figure 18:Server consolidation activity–before configuration ofApril2007 Figure 17:IBM software MLC cost evolution duringthe2006-2011period Figure 16:VWLC SCRTMSUusageanalysis for individualz/OSproducts Figure 15:SCR T extract –peakMSUusagefor z/OSandrelated software products Figure 14:ASC MSUusageprofile – controlled growth asperthe capacity plan Figure 13:ASC soft-capping workload example –R4HAvs. SCRT&IMSUanalysis Figure 12:Unconstrained workload example –R4HAvs.T &IMSUanalysis SCR Figure 11:Workload soft-capping example –R4HAvs. DC Figure 10:SCRTsoft-capping calculation formulaHA orDC –R4 Figure 9:Example Multiple LPAR MSUUsageforT ReportSubmission SCR Figure 8:z/OSviaz196&AWLC Incremental MSUPricingExample Figure 7:MSUPricingStructure –DecliningCost for AdditionalIncremental MSUUsage Figure 6:VWLC Rolling4HourAverage (R4HA)Example Figure 5:VWLC Rolling4HourAverage (R4HA)Formula Figure 4:IBMMainframe Software PricingTimeline Figure 3:CampusModeSynchronous Copy–New Architecture Figure 2:Cross Synchronous Copy–Classical LegacyArchitecture Figure 1:Sample HighAvailability Mainframe Configuration of 66 65 65 65 64 64 64 63 62 61 61 61 60 60 59 58 54 53 53 53 50 49 49 47 46 45 42 42 41 40 40 39 39 29 17 17 16 practices guidelines,whichare publishedaswhitepapers. experience, collaborate onnew technology evaluations anddevelop best providing inwhichmemberscan share aframework knowledge and It isanexecutive andanonprofit peernetwork organization aimed at senior ITinfrastructure managemersfrom majorEuropean organisations. The DTI ALLIANCE(CRiP)isacommunity of CIOs,Chief ITArchitects and Member ofDigital Technology &Innovation organization anddelivers agnostic information. Members are EndUsers only. CRiPisindependentfrom anyvendor orconsulting T in charge ofData Center facilities, Servers, Storage, Desktop, Network, Mobility, (Chief T 170 +companies are CRi P members. Companiesare represented bytheirC lead theirteams withexpanded skills,knowledge, andinnovative agendas. CRiP helpsexecutive members to learn from the experience oftheirpeers, and decision makingandaccelerate timeto market implementation. new practices andtechnologies, andunite thesedynamicelements to improve technologies, processes andevolving trends. They examine key issues, evaluate companies andpublicentitiesgainunparalleled insightinto changing As members ofCRiP,ITInfrastructure &Operations managers ofmajor professionals are partoftheCRiPAlliance SocialNetwork &Community: As ofSeptember 2012,more than1.500ITInfrastructure &Operations elecom, Middleware andDatabases. - CRiPSwitzerlandto becreated in2012 - CRiPLuxembourg created inoct2011 - CRiPFrance created in2007 echnical Officer) orCIO(ChiefInformation officer). CRiP Club desResponsables d’Infrastructure etdeProduction IT T Os usually are T O

PAGE 73 MAINFRAME : IBM zSeries Mainframe Cost Control WHiTEPAPER PAGE 74 - state oftheart collaboration functions -  - to beinformed in«real time»ofITissues - to maintain aneasylinkbetween D CRiP hasdevelopped itsowned private SocialApplication, BeeCRIP, enabling: between members: A SocialNetwork application enablingon-lineactivities their years work. The2012editionattracted 2000+visitors. A 2dayconvention isheldonce ayear. Eachworking group presents theresult of A CRiPonvention isheld once ayear Around 10Themebasedconferences are heldperyear. showcase conference andpublishesawhite paper. opportunities. Meetingsare heldonce permonth.Eachgroup produces atheme’s O themeandincludebothworking andsocialgatheringsto maximizenetworking and optimizedinits20working groups. Working groups focus ononespecificI& The collective wisdomandexperience oftheCRIPmembership isbeingshared and showcases conferences 20 working groups produce white papers business growth. time to market, cost control, aswell asbetter qualityofservices thatencourage Members helpfellow members andtheirorganizations drive changewithbetter and develop best practices. reports. They work collaboratively to stimulate new thinking,compare approaches, This community of1500 managers & experts includestheC Infrastructure &Operation Managers: An executive Communityof1500Plus workgroups meetings “Web 2.0”creation oftrust groups, to share information andideasbetween TI Alliance /CRIPmembers T O andtheirdirect Contacts Club des Responsables d’Infrastructures et de Production [email protected] www.crip-asso.fr

En application de la loi du 11 mars 1957, il est interdit de reproduire ; sous forme de copie, photocopie, reproduction, traduction ou conversion que ce soit mécanique ou électronique, intégralement ou partiellement le présent ouvrage, sur quelque support que ce soit, sans autorisation du CRiP. www.crip-asso.fr 75008 PARIS 15 rue

vignon

Création : fred.lameche - www.anousdejouer.fr