BY BILL CARICO AND WILLEM J. VAN DER ZEL INTEROPERABILITY INSIGHTS OpenEdition MVS: The System, the Strategy, the Significance: Part I — Overview

n the last few months of 1995, Willem van der Zel and I took an “up close and personal” look at OpenEdition MVS. Willem is an old friend and colleague, and is by far the sharpest specialist I have ever met. I went into the project with a lot of preconceived notions, the most Why combine MVS I and UNIX into one prominent due to my skepticism that UNIX would ever play well in the main- frame environment. For example, I was aware that AIX/ESA was only comprehensive system? installed by a half-dozen customers, and that Amdahl’s UTS had only sold Simply because several hundred licenses since 1981. Do the math; it doesn’t take a genius to OpenEdition MVS conclude that UNIX plus mainframes equals limited revenue potential. Well, eliminates UNIX’s imagine my surprise when I heard Amdahl’s president, Joe Zemke’s, proph- shortcomings of esy at Amdahl’s fall 1995 user group meeting: “OpenEdition MVS will dom- , stability, inate the high-end marketplace for the 10 years.” After several months of research, my skepticism turned into support for the strategy. In a and manageability. three–part series, here are Willem’s and my observations about OpenEdition MVS: the system, the strategy, the significance.

UNIX IS HEADED FOR A SHOWDOWN UNIX to lure applications away from MVS The high-end UNIX marketplace have little choice but to brace for the impact. is headed for a showdown. In one corner, OpenEdition MVS neutralizes what was traditional UNIX vendors led by Hewlett previously a primary advantage of UNIX Packard Co. (HP), Sequent Systems Inc., servers — allowing customers access to the AT&T’s GIS division, Digital Equipment plethora of multi-user UNIX applications in Corp. (DEC), Tandem , Inc, and the marketplace. many smaller companies. In the opposite OpenEdition MVS is now competing with corner, IBM’s System/390 servers running the high-end UNIX-server market. It is not UNIX applications on IBM’s flagship MVS being positioned to compete with the low-end . That’s right, IBM S/390 UNIX platforms. Most UNIX multi-user developers have blended MVS and UNIX applications will thrive on OpenEdition into one comprehensive operating system. MVS, while personal productivity applica- This new combo, given the trendy name of tions –- games and countless other single- OpenEdition MVS, is one of the most signif- user programs –- will stay on the desktop. icant software developments to come out of Thus, don’t expect an applications catalog for IBM in recent years. Suppliers claiming to OpenEdition MVS to ever rival the thickness offer mainframe alternatives and promoting of one for desktop computers.

Note: In this series, the acronym MVS, which stands for Multiple Virtual Storage, is used to refer to IBM’s MVS/ESA operating system at its most current level. While this series focuses on OpenEdition MVS, many of the points discussed are also relevant to the OpenEdition VM version of the system, but no attempt will be made herein to highlight any differences.

© 1996, ACTS Corporation, all rights reserved. TECHNICAL SUPPORT APRIL 1996 INTEROPERABILITY INSIGHTS

THE UNIX MARKET a move to compete with UNIX, just to 1975 MVS has been the system the majority The UNIX market covers a broad range re ate an image of openness within [its] of those 20 million users rely on day-in and of systems, from PCs and to installed base.” d ay-out to handle bu s i n e s s - c ritical wo rk . multi-user serve rs. As PCs have become Pip Smith, S e q u e n t ’s senior director During that time, UNIX vendors have made m o re powe r f u l , the distinction between of NT and Major Account Sales, i s n ’t a meager showing in effo rts to displace PCs and workstations has blurred, causing worried either. “If MVS, by meeting Spec MVS systems. growth of the market to slow 1170 can be called UNIX, the next thing Ac c o r ding to the London Tim e s , No vember 1, some. In this series, the use of the term we can expect is for cabbage patches to quali- 1995, 4, a recent study completed by “desktop” systems refers to both PCs and fy as putting greens; if we get the cabbages England market re s e a rch fi rm , C a m b ri d ge personal workstations. close enough, we can call them gra s s ! ” Market Intelligence, estimates that 75 percent Since vendors keep actual sales data confi- Smith said. of all client/server projects are a failure. They dential, determining who is the found that companies often fail to devel- market leader for UNIX is not an op a solid business case for replacing exact science. It depends on their mainframe computers with wh e re the market lines are client/server, and eventually discover the d raw n , whose survey nu m b e rs new system to be far more expensive and are used, and the time period not nearly as effective as the old. being measured. Clearing up the Just ask the telecommunications giant confusion is an undert a k i n g US West, which last year alone spent beyond the scope of this series, $290 million in a failed attempt to con- but a market model provided by ve rt from MVS to Sequent’s UNIX, Intelligence InfoCorp. called /PCX. While re l i abl e ( C I I ) , a market re s e a rch fi rm sources at Sequent said they advised based in La Jolla, Calif., helps to against a conve rsion from MVS to establish a quick reference point. UNIX, US West was not to be deterred, CII defines the low-end UNIX and the project turned out to be a total RISC server market as systems disaster.The easiest-to-port online appli- selling for $25K and less, mid-range between cations were converted, but on UNIX, average $25K and $750K, and high-end as any UNIX- AN UNFAIR FIGHT response times for these transactions bal- based system selling for greater than $750K. Regardless of the rhetoric, IBM’s introduc- looned to five seconds. At the time, only 10 Details of annualized 1995 market- esti- tion of OpenEdition MVS means Sequent, percent of the mainframe workload had been mates, which are updated quarterly, are pro- AT&T GIS (since renamed NCR), H P, c o nve rt e d, and systems management fo r vided for the high-end UNIX market in Figure Tandem, and anyone else selling high-end UNIX took twice as much manpower as was 1. CII estimated that for 1995, a combined UNIX serve rs will soon find themselves needed to manage the remaining 90 percent $5.5 billion would be spent on UNIX systems competing head-to-head with MVS. Even if still running under MVS. In any event, MVS worldwide across these three segments. D av i s o n ’s assessment is accurat e, I B M ’s is alive and well at US West, which is more installed base alone rep resents betwe e n than can be said for the executives overseeing SEQUENT PREDICTS BIG YAWN 40,000 and 50,000 systems wo rl dwide the conversion — they were fired! While such Based on the CII market model, a case s u p p o rting more than 20 million users. could be made that Sequent has the most More importantly, by some estimates these Figure 2: Cost of Computing Summary to lose if OpenEdition MVS is a hit, followed customers spend between 40 percent to 50 by AT&T GIS and HP. Sequent isn’t worried. percent of all IT spending. Due to extensive Hardware, Software, and Maintenance Costs Per Dean Davison, manager of Competitive and cutbacks to their sales force, IBM’s bigger End User Over a Period of Five Years Market Research at Sequent, told ACTS he s h o rt-coming is having too few fe e t - o n - thinks IBM is merely doing everything it t h e - s t reet to find and engage competitors , 30–user system 100–user system can to protect and preserve its mainframe especially since so many vendors sell directly Mainframes $1,920 $1,920 business. He pre d i c t s , “OpenEdition MVS to division and department heads. But when- will be met with a big yawn outside of ever the competition is engaged, head to head, Non–UNIX minis $4,246 $10,452 I B M ’s MVS accounts. IBM is not making it will not be a fair fight. The reason: Since UNIX minis $4,935 $12,337 PCs on LANs $4,060 $9,135

Figure 1: High–End UNIX Market Adding to the above the costs of application software, support, power/space/networking, Other 13.2% Pyramid 10% IBM 6.7% and lost productivity:

30–user system 100–user system

Mainframes $7,964 $7,964 HP 13.2 Non–UNIX minis $14,296 $20,502 UNIX minis $14,985 $22,387 PCs on LANs $15,060 $20,135 Sequent 26.8% AT&T/GIS 15.2% Cray Research 8% Digital 7% Source: Xephon, The Dinosaur Myth, 1995

TECHNICAL SUPPORT APRIL 1996 INTEROPERABILITY INSIGHTS fa i l u res are commonplace at large companies that have tried to The first sign of serious trouble came in the form of limited scala- replace MVS with UNIX or PC/LANs, these projects are quietly bility.The UNIX servers could only support about 10 percent of the abandoned because it is inadv i s able to ack n ow l e d ge mu l t i - m i l l i o n - projected workload (they were able to handle only 10 users per server dollar fa i l u res publ i cly. rather than more than 100 as had been promised by the hardware sup- plier). Second, support costs sky-rocketed. Management grossly under- estimated the incremental support costs associated with maintaining both MVS and UNIX. Take heart UNIX aficionados, A reas such as systems manage m e n t , d i s t ri bution of softwa re, and root cause/pro blem analysis we re all nightmares with UNIX. no longer is painful The re s u l t : O ve rall management costs to synch ro n i ze the new env i- and perilous experimentation ronment added an incremental layer of costs that we re n ’t fo re s e e n and ultimat e ly couldn’t be sustained. The project collapsed under the with a UNIX box necessary weight of its own high costs, but the plug wa s n ’t pulled until the to gain access to UNIX applications. c o m p a ny had invested five ye a rs and more than $400 million try i n g to make it wo rk . OpenEdition MVS is UNIX This costly failure is by no means an isolated occurrence. Other large with MVS under the covers... companies are finding the world of UNIX, Windows/NT, NetWare, and most midrange and PC/LAN systems deployed under the nebulous banners of client/server and open systems to be down right unfriendly when it comes to running businessÐcritical applications, especially in Of course, having a superior product is no guarantee of its success. terms of cost. No matter how wonderful the tech n o l ogy may be, u l t i m at e ly, t h e Ta ke heart UNIX afi c i o n a d o s , no longer is painful and peri l o u s s u ccess of a product rests primarily on marketing clout. For example, ex p e ri m e n t ation with a UNIX box necessary to gain access to Microsoft has many superior and mediocre products, but its marketing UNIX ap p l i c ations. OpenEdition MVS is UNIX with MVS under has leve raged that technical mediocrity to ach i eve market dominance the cove rs , and MVS is ord e rs of magnitude beyond its next in the less sophisticated world of personal computing. competitor in terms of scalab i l i t y, s t ab i l i t y, and ro bustness of At this point, the success or fa i l u re of OpenEdition MVS, as we l l systems manage m e n t . as its Vi rtual Machine (VM) OpenEdition counterp a rt , does not h i n ge on the tech n o l ogy, but rather on IBM’s ability to mount an UNIX PLUS MVS - OVERVIEW OF ADVANTAGES e ffe c t ive marketing campaign and, as stated prev i o u s ly, to engage Many brochures claim a product offers scalability, but the scope and competition head on. Wh at is the upside? If these new systems are meaning of the term depends on the environment. The moral of the p ro p e rly marketed they could easily turn MVS and System/390 into story about the aforementioned financial company teaches us that it is a growth plat fo rm once aga i n . good practice to determine precisely where a system’s practical scaling But IBM needs to intensify its efforts to educate established MVS limits are before plunging headlong into a massive project. and VM customers, as well as new prospects, as to what OpenEdition MVS is well known for its scalability and high performance. It is enhancements can mean to their businesses. able to handle users numbering from the tens to the tens of thousands. Based on ACTS interv i ews with customers , the typical MVS It allows these users to share data, even concurrently update that data, customer lacks an understanding about IBM’s strat egy, is confused and consistently delivers sub-second response time. MVS and S/390 by all of the va riants of UNIX in the marke t p l a c e, d o e s n ’t speak h a rdwa re offe rs the industry ’s most efficient symmetrical mu l t i - the UNIX language, and is not like ly to quick ly grasp the signifi- p rocessing (SMP) env i ronment. Traditional bat ch processing is cance of wh at it means to run UNIX ap p l i c ations on an Open- extremely fast, and has been enhanced further in recent years using Edition plat fo rm . techniques that minimize conventional I/O activity. Up until now, the inability to handle large volumes of I/O has been COMBINING MVS AND UNIX a spoiler for many UNIX projects. Mainframes are designed to handle Why combine MVS and UNIX into one compre h e n s ive system? thousands of I/Os per second, while the biggest UNIX servers,thus far, S i m p ly because OpenEdition MVS eliminates UNIX’s short c o m i n g s only handle hundreds per second. Such limitations explain why of scalab i l i t y, s t ab i l i t y, and manage ab i l i t y. Even though UNIX Tandem plans to add massive I/O capabilities into its server hardware ve n d o rs will never admit it, the attempt to position UNIX as a main in 1996. p l ayer in commercial enterp rise computing has been ex p e n s ive In addition to scalability, MVS offers stability and manageability far and fru s t rat i n g. For ex a m p l e, some consider “UNIX re c ove ry ” to advanced to anything found in the UNIX environment. For instance, be an ox y m o ro n , and the joke going around is that UNIX systems MVS offers the most advanced software error recovery, including sup- m a n agement can be described in one wo rd : reboot. Joking aside, port for dynamic reconfiguration and extended recovery facilities. with ve ry few ex c ep t i o n s , UNIX ve n d o rs simply have not delive re d MVS allows predictable capacity planning, letting upgrades be planned on areas of scalab i l i t y, s t ab i l i t y, and ove rall ability to manage in a non-disruptive manner. the env i ro n m e n t . MVS market dominance has spawned an entire industry of Consider the following true story from a U.S.-based Fortune 100 I n d ependent Softwa re Ve n d o rs (ISVs), o ffe ring systems manage- financial company that requested anonymity: ment tools that minimize support costs and can easily save a customer S eve ral ye a rs ago the company embarked on a major tech n o l ogy h u n d reds of thousands of dollars each ye a r. In add i t i o n , MVS has i n i t i at ive, p ri m a ri ly because it had heard so many claims about the an established base of customers , a large base of proven ap p l i c at i o n a dva n t ages of distri bu t e d / c o o p e rat ive computing. The plan called fo r c o d e, and excellent product documentation. All of this explains the mainframe and PC/LANs to be supplemented with a middle tier why modern mainframes offer the lowest cost per user compared of UNIX serve rs. The business mission was to gain a competitive to UNIX or PC/LAN systems, despite pro p aganda that may cl a i m a dva n t age by developing a wonderful new customer- s e rvice ap p l i c a- o t h e r w i s e. (See Fi g u re 2.) The increase in cost-per-user mov i n g tion that would run under UNIX. f rom a 30- concurrent user system to a 100-concurrent user system INTEROPERABILITY INSIGHTS

operating system from scratch, so UNIX was a logical choice. UNIX History Admittedly, minimum cost also meant limited function, but the concept of an operating system that was in itself a toolbox inspired many pro- UNIX has made a profound impact on the computing industry. To grammers to come to grips with this challenging software, understand understand the significance of what all this means, we have to cast back the concepts, and add function. 25 years... UNIX became the first portable operating system. Additions like edi- When people talk about the origin of UNIX, two individuals are fre- tors, shells, communications, and management utilities, plus the well quently mentioned: known MIT X-Window graphics system, were contributed by different Ken Thompson and Denis Richie. These gentlemen are widely recog- individuals and organizations. Furthermore, the strategy to donate the nized as being instrumental system to universities paid off by providing a ready supply of techni- in creating UNIX along with others working in the Computer Science cians familiar with the environment. Research Group at Bell Telephone Laboratories in 1968. UNIX was Error recovery for UNIX originally was primitive or non-existent, and derived from Multics, an operating system designed for the brave and large organization, searching through the to allow multiple users to run simultaneously while allowing them to share was often the final arbiter. Errors, misbehavior, and bugs were thus the their data with each other. cause of much of the UNIX criticism. For example, it certainly is not Multics was developed jointly by General Electric and Massachusetts acceptable for the screen to continually scroll: Institute of Technology (MIT). The intent of creating UNIX from Multics was to produce a scaled–down, reduced function operating system that I am not a typewriter could provide base function for a single–user system. This helps to I am not a typewriter explain why those companies that later picked up UNIX needed to sup- I am not a typewriter plement, or rebuild, function back in to make UNIX a more sophisticated just because you have tried to load a data file which exceeds the allowed system that could handle multiple concurrent users. maximum size. (Observed on AIX 2.2.1 for the RT PC.) To better understand UNIX’s , it helps to know that Improvements in reliability and stability have been added over time UNIX was originally designed for internal use by friends, co-workers, via extensions (adding new code) to the system. Since its beginnings in and other highly computer-literate people who could not type. Anyone 1974, Tandem has tried to make its UNIX fault tolerant. Tandem has added who ever had to work with one of those old clunky terminals appreciated extensive recovery logic and guarantees data integrity, calling its UNIX a system that kept keystroke requirements to an absolute minimum. Non-Stop Kernel (NSK). Similarly, AT&T GIS (since renamed NCR) has That is why still today, the UNIX operating system has many short cryp- added extensive recovery logic and calls the system UNIX-MPRAS, tic commands that one can string together. where RAS stands for Reliability,Ava i l a b i l i t y , and Servi c e a b i l i t y . But most A later port to a 16-bit (PDP-11) was done with the variants of UNIX suffer serious availability shortcomings, though most aid of a language specially designed for the project called B. Denis have been refurbished through the years to do multitasking and to a large Richie improved and refined B calling it C. UNIX was then re-written in degree shed the single-user nature of UNIX. C, support for multitasking was added, and it became relatively easy to Security, long a neglected item, was given detailed attention and th e port UNIX to other machines. In the mid-1970s made UNIX modern versions of UNIX out there, like IBM’s AIX version 4, have plugged available to universities for free. In 1979, using money supplied by the all the back-door entries. Countless known bugs have been fixed, assist- U.S. government, the University of California at Berkeley began ad d i n g ed by millions of users who have been very involved and useful on the functionality to UNIX and subsequently released the Berkeley Systems networks, and prodded on by highly effective worms and other infection Distribution (BSD) version of UNIX, which added many subtle changes programs. and improvements. Separately, Bell Labs UNIX has now primarily become a network-oriented operating sys- continued to develop its version of UNIX, resulting in two major varia- tem and TCP/IP has become a de facto industry standard. Au s t r a l i a n tions of UNIX with similar functionality, but different code. From these passport and visa applications world-wide run on a network of UNIX com- two mainstream versions (AT&T and BSD), many other UNIX systems puters. Additions, such as and were cobbled together and marketed under a plethora of operating system X-Windows, greatly enhance where a UNIX system can be deployed. Fo r names usually including the letter X (e.g., , AIX, SCO-UNIX, HP- example, the overseas branch of a large organization can in many UX). Since the UNIX name is trademarked, variants were renamed to different ways and even display graphics locally from an application run- avoid trademark hassles. ning overseas.

UNIX Strengths and Limitations Today, an operating system based on UNIX is available on nearly UNIX System V - The Itinerant Operating System every computer hardware product. C and C++ are the most common IEEE, COSE, and X/Open all have been busy trying to fix UNIX’s languages for UNIX and other platform software development. Also, portability problems. As stated previously, AT&T sold its UNIX System many different user interfaces to UNIX, called shells, have been promul- Laboratories, along with UNIX System V, to in 1993. Soon after, gated, the most popular C-shell was added with the Bourne and Korn Novell gave the official UNIX branding rights to X/Open. This move was shells (there are many, many more shells: i.e., bash, tsh). UNIX was primarily to alleviate fears that UNIX might not receive adequate at t e n - used extensively in the U.S. Department of Defense Advanced Research tion and development in the hands of Novell. Prior to this, X/Open was Agency and thus formed an integral part of the beginnings of the noted for creating the Common Application Environment (CAE) specifi- Internet. So the result was a quick and inexpensive multi-user and multi- cation which was 100 percent product independent. CAE has been since tasking operating system easy to port to different hardware platforms. The renamed Single UNIX Specification, and X/Open is no longer product so u r ce code was available for about $20,000 (free to universities), making independent, despite any claims to the contrary. With extensive UNIX UNIX the best price/performance operating system. Sun, for example, involvement their objectivity is certainly questionable, and it should sur- needed an operating system that was economical in terms of price to prise no one that X/Open views UNIX as “the leading open systems com- run on their workstations. It was cost prohibitive to build their own puter operating system.”

TECHNICAL SUPPORT APRIL 1996 INTEROPERABILITY INSIGHTS

UNIX at IBM Through the years IBM has ported UNIX to nearly all of its computers, from the smallest to the largest. PC/IX was ported to the as early as January 1984. was later ported to the IBM Series/1. IBM combined AT&T System V UNIX with the Berkeley System Distribution (BSD) version of UNIX, and added its own system management, journaled file system, and other enhancements, to create its Advanced Interactive eXecutive (AIX) operating system. AIX ran on Bill Carico has more than 20 years of experience in the computer industry. personal computers (IBM’s PC R/T), and later workstations (IBM’s RISC He is co–founder and president of ACTS Corporation, which specializes in System/6000 series of computers). Up consulting and education. Bill will be presenting at NaSTEC ’96, which will to this point, AIX has been IBM’s most successful UNIX. IBM held in Orlando, Fla., May 5 – 8, 1996. advertisements boast that there are now more than 10,000 applications marketed for their family of UNIX-based RISC System/6000 servers. Willem J. van der Zel is a recognized authority on UNIX, especially in the The latest version of AIX runs on the IBM family of PowerPC-proces- graphics arena. He also has several years of experience working with MVS. sor-based client and server platforms. UNIX variants are Before becoming an independent consultant in 1994, Willem worked for IBM usually tied to one particular hardware architecture, but now Sun’s for 13 years, as a systems engineer and in various technical support roles. Solaris is also being offered for certain models of the PowerPC. Se l d o m do different variants of UNIX run on the same hardware. ts ©1996 Technical Enterprises,Inc. Reprinted with permission of TechnicalSupportmagazine. For subscription information,email [email protected] or call 414-768-8000,Ext. 116. exposes the ve ry serious scalability limitations of mini-computer and PC/LAN systems. Now that the stage has been set,Part II will examine IBM’s strategy, predict what its impact will be,and provide an overview of the capa- bilities of the system. ts

TECHNICAL SUPPORTAPRIL 1996