Compactpci and Advancedtca Systems / October 2006

RSC# @ www.compactpci-systems.com/rsc RSC# @ www.compactpci-systems.com/rsc ® Volume 10 • Numb e r 8 OCTOBER 2006 CompactPCI www.compactpci-systems.com ® www.advancedtca-systems.com and AdvancedTCA Systems The Magazine for Developers of Open Communication, Industrial, and Rugged Systems

COLUMNS FEATURES 8 Editor’s Foreword GUEST: TELECOM SOFTWARE The universe is expanding 22 How DSO, COTS, and open architectures can help By Joe Pavlat solve the ®looming telecom software crunch 10 Software Corner By Michael Christofferson, Enea Embedded Technology Pioneering model driven development By Curt Schwaderer Compac tPCI SPECIAL: MEDIA SERVERS AND SOFTSWITCHES 14 Specification Corner 26 I-TDM: Supporting TDM voice in the age of PICMG specification update ® By Rob Davidson MicroTCA and AdvancedTCA and By Robbie Dhillon, Accolade Technology, Ian MacMillan, Interphase, 16 Technology in Europe and Amir Zmora, Surf Communication Solutions Powered by CompactPCI By HermannA StrassdvancedTCA32 SelectingSystems a modular media gateway to enable 18 Technology Update VoIP and other content-rich media services Advanced traffic management aids converged IMS applications By Venkataraman “VP” Prasannan, RadiSys By Peter Yan TECHNOLOGY: LIQUID COOLING

36 Liquid-cooled embedded computing initiative By Tahir Cader, Eric Grabowski, Joe Pavlat, and John Peters, EVENTS Liquid-Cooled Embedded Computing Initiative AdvancedTCA Summit October 17-19 PRODUCT GUIDE: MicroTCA Santa Clara, CA 38 Clocking in: Real-world MicroTCA needs close www.advancedtcasummit.com clocking/fabric interaction By Will Chu, CorEdge Networks SDR Forum Technical Conference November 13-17 42 MicroTCA – a new standard for the battlefield Orlando, FL By Rob Persons, Motorola Embedded Communications Computing www.sdrforum.org 44 MicroTCA offers a direct solution for tight cost and size restraint applications COVER: By Stuart Jamieson, Emerson Network Power A general purpose computing and software development platform, a WiMAX basestation, a 3G/4G basestation, and a quadruple/triple play IP Multimedia System (IMS) platform are among the working MicroTCA systems CorEdge Networks is developing, with particular attention to clock E-LETTER and fabric interaction. See page 38. October: www.compactpci-systems.com/eletter MicroTCA: Modular and scalable MicroTCA Carrier Hub (MCH) photo courtesy By Volker Haag, Schroff CorEdge Networks. November: www.compactpci-systems.com/eletter A top-down perspective of IMS, AdvancedTCA, and blade servers By George Kontopidis, NMS Communications Published by:

© 2006 OpenSystems Publishing ® CompactPCI, PICMG, AdvancedTCA, ATCA, MicroTCA, and their logos are registered trademarks of the WEB RESOURCES PCI Industrial Computer Manufacturers Group. Subscribe to the magazine or E-letter at: TM AdvancedMC and CompactTCA are trademarks of the PCI Industrial Computer www.opensystems-publishing.com/subscriptions Manufacturers Group. © 2006 CompactPCI and AdvancedTCAOpenSystem Systems s Industry news: All registered brands and trademarksPublishing in CompactPCI and AdvancedTCA™ Systems are property of their Read: www.compactpci-systems.com/news respective owners. Submit: www.opensystems-publishing.com/news/submit Submit new products at: / CompactPCI and AdvancedTCA Systems / October 2006 www.opensystems-publishing.com/vendors/submissions/np

RSC# @ www.compactpci-systems.com/rsc CompactPCI ® ® and AdvancedTCA Systems

OpenSystems Publishing™ OpenSystems Publishing CompactPCI ® Advertising/Business Office Regional Sales and AdvancedTCA Systems 30233 Jefferson Avenue Jane Hayward St. Clair Shores, MI 48082 Regional Manager – California Tel: 586-415-6500 n Fax: 586-415-4882 [email protected] A n O p e n S y s t e m s P u b l i c a t i o n Vice President Marketing & Sales Phil Arndt Patrick Hopper Regional Manager – East Coast Communications Group [email protected] [email protected] n CompactPCI and AdvancedTCA Systems Business Manager Richard Ayer Karen Layman Regional Manager – West Coast n CompactPCI and AdvancedTCA Resource Guide [email protected] n CompactPCI and AdvancedTCA E-letter

Editorial Director Joe Pavlat Sales Group [email protected] International Sales Dennis Doyle Assistant Managing Editor Anne Fisher Senior Account Manager Stefan Baginski [email protected] [email protected] European Bureau Chief Senior Editor (columns) Terri Thorson [email protected] Tom Varcie [email protected] Account Manager Dan Aronovic Technology Editor Curt Schwaderer [email protected] Account Manager – Israel [email protected] [email protected] Doug Cordier Associate Editor Jennifer Hesse Account Manager [email protected] [email protected] Reprints and PDFs European Representative Hermann Strass Barbara Quinlan [email protected] Account Manager [email protected] Call the sales office: 586-415-6500 Art Director Steph Sweet Andrea Stabile Senior Web Developer Konrad Witte Advertising/Marketing Coordinator Graphic Specialist David Diomede [email protected] Circulation/Office Manager Phyllis Thompson Christine Long [email protected] E-marketing Manager [email protected] OpenSystems Publishing Editorial/Production office: 16872 E. Ave of the Fountains, Ste 203 FountainOpenSystem Hills, AZ 85268s Tel: 480-967-5581Publishing™ n Fax: 480-837-6466 Website: www.opensystems-publishing.com

Publishers John Black, Michael Hopper, Wayne Kristoff Vice President Editorial Rosemary Kristoff

Embedded and Test & Analysis Group Editorial Director Jerry Gipper Editorial Director Don Dingee Technical Editor Chad Lumsden Associate Editor Jennifer Hesse Special Projects Editor Bob Stasonis European Representative Hermann Strass

Military & Aerospace Group Group Editorial Director Chris Ciufo Assistant Editor Sharon Schnakenburg Senior Editor (columns) Terri Thorson European Representative Hermann Strass European Bureau Chief Stefan Baginski

CompactPCI and AdvancedTCA Systems (USPS 019-288) is published nine times a year (Jan, Feb/March, April, May, June, July/Aug, Sept, Oct, Nov/Dec) by OpenSystems Publishing LLC, 30233 Jefferson Avenue, St. Clair Shores, MI 48082. Print ISSN 1098-7622, Online ISSN 1550-0381. CompactPCI and AdvancedTCA Systems is free to qualified engineers or management dealing with or considering open system technologies. For others, paid subscription rates inside the US and Canada are $63/year. For first class delivery outside the US and Canada, subscriptions are $90/year (advance payment in US funds required). Periodicals postage paid at St. Clair Shores, MI, and at additional mailing offices. Canada: Publication agreement #40048627. Return undeliverable Canadian addresses to: WDS, Station A, PO Box 54, Windsor, ON N9A 615. POSTMASTER: Send address changes to CompactPCI and AdvancedTCA Systems 16872 E. Ave of the Fountains, Ste 203, Fountain Hills, AZ 85268 RSC# @ www.compactpci-systems.com/rsc

/ CompactPCI and AdvancedTCA Systems / October 2006 RSC# @ www.compactpci-systems.com/rsc Editor’s Foreword

By Joe Pavlat CompactPCI & AdvancedTCA Systems The universe is expanding

The universe of new platform standards types of AdvancedMCs with MicroTCA for communications and telecom appli- Carrier Hubs (MCHs) that support cations continues to grow in depth and specific networking protocols and clock “The shift towards the breadth. This issue digs deeply into a types, developers can leverage the variety of related topics. Certainly one MicroTCA platform to deploy a wide integrated battlefield of the most discussed topics of late is array of applications. MicroTCA, and here you’ll find insiders’ and net-centric perspectives on this new standard, ratified Lest one think that MicroTCA is just for by PICMG in July. telecom applications, Rob Persons from warfare has created Motorola shows us how MicroTCA is Work on the first open architecture being applied to modern military applica- a need for the same specifically designed for demanding tions. The shift towards the integrated bat- high performance telecom applications, tlefield andnet-centric warfare has created high performance, AdvancedTCA, began in 2001. More than a need for the same high performance, IP- 100 PICMG member companies partici- based systems that are typical of the tele- IP-based systems pated, and the standard was ratified at the com world. It is expected that PICMG will end of 2002. Next came the challenge to begin work to formalize a military version that are typical of maximize AdvancedTCA flexibility and of MicroTCA later this year. modularity. A new hot pluggable, man- the telecom world.” aged, fabric-based mezzanine card would In the September issue of CompactPCI be needed, the Advanced Mezzanine Card. and AdvancedTCA Systems, the work Ratified in early 2005, the AdvancedMC being done by the Liquid-Cooled Embed- exceed those of hardware these days, and specification quickly generated momen- ded Computing (LCEC) initiative was the rapid expansion of code isn’t being tum. As many as eight AdvancedMCs can introduced; in a follow-on article, which matched by a commensurate increase in plug onto an AdvancedTCA carrier card, begins in this issue and is available in the number of programmers. He explains allowing system designers and integrators full at www.advancedtca-systems.com, how open architectures in both hardware to easily customize a common base the LCEC’s two-pronged development and software help address this problem, platform configuration. AdvancedMC approach of a Static Display Model and and introduces us to Device Software cards are very powerful, and it wasn’t a prototype system, developed under Optimization, or DSO. long before system architects began to a Defense Microelectronics Activity think about plugging them directly onto (DMEA) program, is detailed. Also addressing the topic of new software a backplane in a small chassis. Doing so development methods, Curt Schwaderer would create a physically small, low cost Time Division Multiplexing, or TDM, has writes in the Software Corner about a new platform with very high performance and been the basis of digital voice transmis- concept called Model Driven Develop- scalability. And so MicroTCA was born. sion and handling for a long time. In the ment. Curt takes us through high-level con- new world of IP-based communications, cepts and explains how they fit together Contributors a method is needed to support legacy to create a new software development Stuart Jamieson of Emerson Network TDM data over IP networks. Internal environment. Power provides an excellent introduction TDM (I-TDM) is a new standard devel- to the MicroTCA architecture and its oped by PICMG to support TDM traf- Things have been very busy at PICMG applications. Stuart is a true expert on fic on platforms such as AdvancedTCA lately, and PICMG’s Rob Davidson gives MicroTCA, as he served as document and MicroTCA. Three industry experts, us an update on what the organization has editor during the specification develop- Robbie Dhillon, Ian MacMillan, and been up to of late, which includes ongo- ment. Stuart and his colleagues also pro- Amir Zmora team to provide us with a ing updates and enhancements to existing vided significant technical contributions tutorial about I-TDM, how it works, and specifications, as well as new initiatives. and guidance to the standard. how it is applied. This issue is an essential one for catching Volker Haag from Schroff provides Venkataraman Prasannan from RadiSys up on a quickly expanding universe as we additional details about MicroTCA in also explores issues of handling voice wind up the year. Read on! this month’s E-letter (www.compactpci- over the new networks. He describes how systems.com/eletter). Schroff was also a key AdvancedTCA and MicroTCA can be Joe Pavlat player in the development of MicroTCA. used to handle Voice over IP traffic.

Will Chu of CorEdge Networks explains Michael Christofferson from Enea ex- that by intelligently pairing specific plains how software development costs Editorial Director

/ CompactPCI and AdvancedTCA Systems / October 2006 RSC# @ www.compactpci-systems.com/rsc Software Corner

By Curt Schwaderer CompactPCI & AdvancedTCA Systems Pioneering model driven development

Since software and systems engineering has (At www.prismtech.com a comprehen- as Data Distribution Service (DDS) and been around, so too has systems and soft- sive white paper is available that covers distributed object middleware such as ware modeling. Modeling delivers benefits the details of the MDD approach.) CORBA. that include proving out the system prior to detailed design and implementation so What is model driven MDD hurdles that errors can be detected and eliminated development? Software and system modeling has been earlier in the development cycle. MDD is a development practice where around for many years. Typically the mod- high-level, agile, and iterative software eling occurs at the proof-of-concept stage Witnessing the operational aspects of models (often domain-specific) are cre- using third generation languages such system requirements is also beneficial. ated and evolved as software design and as C/C++ or Java. Universal Modeling Modeling provides a proof-of-concept implementation takes place. The key Language (UML) tools have been refined that saves time and money during the defining characteristic of MDD is that over time. Use of UML has spanned development cycle by facilitating the the model literally becomes part of the anywhere from modeling to being used collective agreement of all stakeholders development process. Contrast this with to develop designs. However, until this pertaining to the requirements and func- an approach such as the waterfall devel- point, UML has historically lacked tional operation of the system. opment process where modeling appears semantics to describe the system at the as a separate step in the process and tends business or embedded application level. However, modeling has historically been to get left behind once the development Work must be done to define the appli- disjoint from the development cycle. Often proceeds to the next phase. cation within its domain, with a separate the model is discarded once the concept step required to translate that description has been proven, as was the situation with Work in defining model driven develop- to UML. Software frameworks, libraries, CASE tools in the 1990s, and development ment best practice is ongoing by various and software components available on the begins from scratch. This chasm between industry groups, including the Object Web in open source form have also been modeling and the development process Management Group (OMG), an organi- used for modeling. leads to design flaws and an implementa- zation of end users and software vendors tion that does not behave according to the for developing industry standards for The biggest obstacle to achieving MDD original model. For many years developers the software life cycle. The idea behind with these traditional tools and compo- have been striving to make the model MDD is to model an application and use nents is domain specificity. Application a more integral and iterative part of the a single repository where the high-level stakeholders need to describe the system process. This approach would enable the model of that application (and its systems requirements and functionality in the model to serve as the design and implemen- environment) is maintained. This enables domain of that system, including: tation starting point for the development. non-software-engineering stakeholders What’s more, the model could be evolved as (for example systems designers and engi- n Information that the system keeps the development proceeds. In this column, neers) to maintain control (well past the n Behavior of the components in the we will explore a development approach initial modeling stage) of requirements system called Model Driven Development or and functionality for the system during n Availability of the information to MDD. The evolution of MDD has been development. Thus, MDD allows business those components going on for some time, but recent mod- and technical personnel to help define eling and development tool innovations and maintain the model in a very real and The semantics of UML or third genera- are making MDD a viable approach to meaningful way, resulting in systems and tion languages such as XML, C/C++, and software and systems engineering. applications that more accurately satisfy Java, should not limit these descriptions. requirements on initial deployment. Nothing is worse than to have an experi- Proliferating MDD throughout the enced stakeholder in the domain unable embedded space MDD’s component architecture together to express to the engineering team impor- PrismTech, a company whose core com- with automatic code generation (latter- tant aspects of the system because the pro- petency has historically been software day interpreters between the model and gramming language puts up obstacles. tools and middleware, believes now is the source and unit test code) and high- the time to make MDD a reality. The performance, low-overhead middleware A chasm company embraces the proliferation of meet the needs of networked enterprise So, there is a chasm that exists between the MDD throughout the embedded systems and embedded systems where application architecture created by the domain experts space. PrismTech sees MDD as useful functionality is distributed among net- and third generation languages, tools, for enterprise, net-centric, and mission- work nodes. frameworks, and other components used critical applications. They have suc- to implement it. In this chasm, domain cessfully used the approach to develop As a result, MDD is highly compatible specific requirements decouple from the real-world embedded software products. with publish/subscriber middleware, such technical design and implementation of

10 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 11 @ www.compactpci-systems.com/rscCompactPCI and AdvancedTCA Systems / October 2006 / 11 Software Corner the system. This decoupling can lead to that transform the domain specific model to general purpose platforms with soft- derailment of the project if a system does into executable software for a specific ware that implements the functionality not function as intended. enterprise or embedded platform or prod- of the radio. This enables radio systems uct family. Table 1 describes the steps to become easier to maintain, lengthens MDD bridges this chasm using Domain involved in model driven development. the deployment life cycle, and reuses or Specific Modeling Language (DSML). enhances existing software algorithms. Those familiar with the Eclipse Modeling Each of the steps for model driven devel- environment will recognize that DSML opment is accounted for in the OpenSplice Development of SDR starts with the brings high-level engineering/modeling product line. Using OpenSplice, the Software Communications Architecture work and low-level implementation pro- model produces a collection of develop- (SCA). SCA is an SDR model created gramming together as two well-integrated ment artifacts that are used to generate in 1999 by a consortium of leading parts of the same job. the implementation. military radio developers. SCA isolates abstractions and describes how they work The benefits of the MDD approach are A Software-Defined Radio together within the domain of an SDR. numerous. There is a significant increase using MDD in cross discipline collaboration. The exis- PrismTech has developed a Software- The next step is to create a formalized tence of the DSML means nontechnical Defined Radio (SDR) product using grammar, or Domain Specific Language and technical participants all speak the MDD and the DDS tools. To drive home (DSL). Most SCA implementations same language without getting into tech- the impact of MDD, we will overview use third generation languages applied nical programming details that may cause some of the key points of the PrismTech directly from the SCA specification. The miscommunication or errors in interpre- SDR product so you can have a look at PrismTech approach raises the level of tation of the system requirements. Learn- MDD in action. abstraction to define formalized meta- ing curves associated with this approach model components that are expressed in involve understanding of the domain Like many other areas, radio technology terms of a language workbench, specifi- specific terminology, not details behind is moving from custom hardware systems cally, the Eclipse Modeling Framework. programming languages and tools.

Model driven development and domain specific languages So, how is model driven development done? First, a DSML must be defined for the application, forming a foundation for communication among all stakeholders. Next, a domain specific editor captures requirements and high-level operation of the system. Finally, transformation engines that take the output of the domain specific editor and generate source and/or link to functional components that can be built into the system executable complete the model driven development environment.

Using model driven development, modeling and programming activities blend into the same development process. The models become direct input into the development process, not just an aid to design activity. Domain specific models are developed to be machine processed for integration with implementation generators. Further, implementation of the Figure 1 model is done from the design perspective, not according to third generation General MDD Methodology PrismTech OpenSplice Process language semantics. Isolate the abstractions and how they work together The Data Distribution Service Specification PrismTech is using tools and middleware Create a formalized grammar for these – DSL Create a formalized DSS meta-model that implement DDS concepts and pro- Create a graphical representation of the grammar – DSGL Create a DDS-specific graphical tool cesses in its OpenSplice product family Provide domain specific constraints – DSGL, DSCL Program the constraints into the tool (Figure 1). OpenSplice is targeted to aid in the development of domain specific Attach generators for necessary transformations C/C++ and Java generators languages. It also allows for generators Table 1

12 / CompactPCI and AdvancedTCA Systems / October 2006 Once the DSL is completed, a meta-model a high level of correctness with pre- process has more far-reaching implica- for SCA is specified. The meta-model is validated logic and the software arti- tions than one example in one domain. the key ingredient that allows end users facts derived directly from the model. The OpenSplice MDD/DSS environment with knowledge in the domain to specify The SCA architecture is captured in the is capable of extending model driven what they want to have, not how it is done. PrismTech meta-model, lowering the cost development into a number of areas. Sys- The meta-model also allows the end user of entry significantly for companies wish- tems and software continue to become to directly affect model implementation. ing to move to SDR. With the MDD para- increasingly complex. Now more than digm, it’s possible to determine defects in ever, it’s important to leverage domain Next, a graphical description of the gram- the system at modeling time rather than expertise into the heart of the develop- mar is created, called Domain Specific during system integration. This approach ment process. Model driven development Graphical Language (DSGL) and Domain helps achieve architectural consistency and products make this transition possible Specific Views (DSV). The PrismTech across the model and implementation. in an efficient and effective way. Spectra SDR PowerTool modeling environment allows the end user to define Conclusion For more information, contact functional components and connect them This SDR example covers specifics in Curt at cschwaderer@opensystems- together to specify processing character- the SDR domain. However, the MDD publishing.com. istics. All of this is done using the DSL.

The language workbench’s programming facilities implement, within the domain specific graphical language, the Domain Specific Constraint Language (DSCL). So, the programming within the language workbench includes a structural framework for the rules of the domain. When end users develop their components and systems, the constraint language ensures that they are conforming to the domain requirements. The constraint language is an important part of maintaining the correctness of the model and requires domain experience.

Ultimately, the DSL is transformed into an executable format. This occurs through Domain Specific Generators (DSG). For embedded systems, PrismTech notes these generators may have multiple tar- gets on the platform, including FPGAs, general purpose processors, or digital signal processors. As a result, the DSGL tool needs to be able to iterate over the model, interacting with multiple domain specific code generators to produce multiple types of executable code.

PrismTech’s OpenSplice DGSL tool can declare a component in the SDR domain. This tool can also create software artifacts from the DSGs, keep code coverage information, and keep test case generation information for the components within the model. I was impressed by his tool chains’ ability to bridge the gap between a graphical model component and its VHDL description to C source code with test case and code coverage information.

PrismTech notes a number of advantages of the OpenSplice MDD paradigm in the modeling and development of SDR versus using third generation languages. Domain experts can be far more productive in the MDD environment due to the abstraction. The domain specific generators provide RSC# 13 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 13 Specification Corner

By Rob Davidson CompactPCI & AdvancedTCA Systems PICMG specification update

It has been another busy year for those in version of the AdvancedTCA base speci- SCOPE Alliance for telecom platforms. the PICMG who, for reasons best known fication is PICMG 3.0 R2.0, denoting the RES has set ambitious goals but has a to themselves, love to delve into the second revision. ECNs are effectively strong membership working hard, and they ultimate in fine print of PICMG speci- addenda to the specification. They are will have an important, long-term impact. fications. These spec heads have been used to address relatively minor changes creating entirely new documents that the that do not require a reprinting of the New fabric definitions industry has been eagerly awaiting, as well entire specification. Revisions are more A new fabric definition for AdvancedTCA as updating existing popular specifications extensive and incorporate previous ECNs is well into member review. PICMG 3.6 to incorporate the latest developments. into a new document. PICMG is always defines a fabric called Packet Routing very careful to ensure that both ECNs Switch (PRS) for rapid transport of pack- Achievements and revisions have the smallest impact on ets over an AdvancedTCA backplane. PICMG’s biggest achievement of the backward compatibility possible. year so far is the July release of the AdvancedMC will soon be getting two much-anticipated MicroTCA specifica- This year has seen ECN 002 for PICMG completed fabric definitions. PICMG tion (PICMG MTCA.0 R0). MicroTCA 3.0 R2.0 (AdvancedTCA) adopted. This AMC.2 defines an Ethernet fabric, while completes the trio of specifications that ECN incorporates two years of learning PICMG AMC.4 will define Serial RapidIO includes AdvancedTCA and AdvancedMC. from the industry and includes clearer lan- mapping. This trio covers the wide range of equip- guage that removes ambiguities, as well ment in the telecom network from the core as many technological updates. PICMG Other updates underway include a revi- to the edge. Even though products com- AMC.0 R1.0 (AdvancedMC) also had sion to AdvancedTCA, PICMG 3.0 R3.0, plying with these specifications are used ECN 001 issued in June. In addition to which will incorporate the recent ECNs in many markets, the Telecom Computing including improvements similar to those and other updates. A second ECN for Architecture (the TCA in AdvancedTCA) of AdvancedTCA, this ECN added new AdvancedMC is also underway. framed the design goals. carrier connector types. The future MicroTCA takes the overall framework Work in progress This column has reported on the progress of AdvancedTCA, uses the module form Meanwhile, work continues on several of subcommittees that have been formally factor from AdvancedMC, and defines other projects within PICMG. This approved by the PICMG Executive Board. small form factor system architecture. includes AdvancedTCA300, a variant of Other efforts go on in the background of MicroTCA systems scale from small, AdvancedTCA designed to fit in 300 mm PICMG that will probably turn into formal inexpensive simplex implementations to deep equipment cabinets that are used by work at some time. These background fully redundant 6-nines high availability some carriers. efforts will be the focus of much attention systems. As the AdvancedMC cards are in the coming year. As they are not formal- the plug-in components of the system and The PICMG Requirements Engineering ized yet, I cannot report on them here. You most are already available, MicroTCA Subcommittee (RES) has major work will need to join PICMG to participate. should be off to a strong start. Even though underway, but will not deliver a specifica- Hint: 10 Gigabit Ethernet. the form factor is small, the specification tion. Rather, this committee is working is not – at 2.4 lbs and 538 pages it rep- on new practices that will make PICMG Visit www.picmg.org/specifications.stm resents substantial detailed work by the specifications easier to use and improve for all PICMG specification updates. subcommittee that has labored diligently the interoperability of products devel- over the last year and a half. oped to them. For example, the first task Rob Davidson has been vice president the RES has set itself is to enumerate the of marketing for PICMG since 1995. He After the release AdvancedTCA specification and fully cata- has held senior marketing positions at Once a PICMG specification is released log the optional language. This will make it Intel and Ziatech and is now an inde- the work does not go away. Issues are easy to clearly reference the section of the pendent consultant based in San Luis identified as companies develop products specification when identifying which fea- Obispo, California. and test them at the PICMG Interoperabil- tures and options have been implemented ity Workshops, and new technologies can in a product. The practice they develop in For more information, contact Rob at: impact the specifications as new ways of this exercise will then be applied to future solving problems are identified. PICMG specifications as part of the development PICMG has two methods of updating specifica- process. Once this exercise is complete the 1334 Oceanaire Drive tions, which include the revision and the RES will develop a standardized means of San Luis Obispo, CA 93405 Engineering Change Notification (ECN). mapping profiles for vertical applications Tel: 805-542-0999 PICMG specifications are referred to by into the specification. The first profile to E-mail: [email protected] their number and revision. The current be mapped will the one developed by the Website: www.picmg.org

14 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 15 @ www.compactpci-systems.com/rsc Technology in Europe

By Hermann Strass CompactPCI & AdvancedTCA Systems Powered by CompactPCI

Providing power at all times by IEEE as a Technical Report (TR 1550) in 1999. The IEEE In order to make it easy for power companies to control the and IEC agreed in the mid 1990s to use this TR, also known as flow of electrical power across national boundaries (there are UCA 2.0, as one of the major inputs to what became IEC 61850, many boundaries in Europe), the International Electrotechnical a global standard composed of 10 parts. IEC 61850 uses logic Commission (IEC) has generated the IEC 61850 standard. Power models for components and protocol-agnostic real-time commu- companies and energy traders use equipment that adheres to this nication services. It uses TCP/IP and XML for communication standard to communicate and interoperate with others who also and ISO 9506, the Manufacturing Messaging Specification, as use such standardized equipment. an application.

The international standard IEC 61850, Communication Networks The IEC 61850 standardization committees aimed to foster and Systems in Substations, was ratified in 2002 with the first additional standards and side standards under the umbrella of products appearing in 2004. It is supported by major utility IEC 61850. Several of these, including hydroelectric power, wind companies such as AEP, EdF, and EON as well as major power power, and distributed energy resources have been implemented. control or automation companies such as ABB, Alstom, GE, The gas and water works industry in the United States, includ- and Siemens. Efforts in standardization started in 1987 in the ing the American Water Works Association (AWWA) and others, United States and in 1988 in Germany. In the United States, the has accepted and is promoting the use of IEC 61850. Given that Electric Power Research Institute (EPRI) developed a Universal IEC 61850 is neither application nor product specific it can also Communications Architecture (UCA), which was also published be used for other process control and automation applications

RSC# 16 @ www.compactpci-systems.com/rsc

16 / CompactPCI and AdvancedTCA Systems / October 2006 outside the electrical utility area. The IEC 61850 motto is “One World, One Technology, One Standard.”

In order to control electrical power on its way from generator to consumers in factories or homes and across national boundaries, ABB Power Technologies (Sweden) uses Intelligent Electronic Devices (IEDs). The ABB IED670 family of products satisfies the stringent requirements of the IEC 61850 standard in providing extensive communication capabilities and interfaces for maximum compatibility. The extensive I/O capability on analog and binary I/O allows implementation of the most advanced applications. Utilization of common hardware and software components simplifies setting, commissioning, and maintenance. IED670 features extensive logic capabilities that can be used, for instance, to perform load transfer, automatic disconnector open- ing, and other functions. The ready-to-use IED670 packages are designed to achieve optimum operation of the power system. The packages are delivered preconfigured and with default settings for line distance, line differential, transformer protection, and bay control applications.

The IED670 products from ABB are based on CompactPCI boards from Kontron (Germany). Kontron’s CP620 board was modified into the CP6200 board to comply with ABB’s EMC requirements and I/O capabilities in accordance with the IEC 61850 standard (Figure 1). The CP6200 is based on a PowerPC processor running under the VxWorks real-time operating system, a processor/software combination, which is most frequently used in such demanding applications. The CP6200 is equipped with one PMC and two PC•MIP mezzanine sockets for I/O flexibility through rear I/O connections. IEC 61850 uses Ethernet and TCP/IP, making it no surprise that the CP6200 is equipped with two full-duplex fast Ethernet channels. For reliability reasons, memory is protected with Error Correction Code (ECC).

Figure 1

IEC 61850 was made for a very broad and varied application range. ABB claims that they were the first in the market to have products that meet all the requirements of IEC 61850. The IED670 family of products based on Kontron’s CP6200 CompactPCI boards is a steppingstone into a large market.

For more information, contact Hermann at [email protected]. RSC# 17 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 17 Technology Update

By Peter Yan CompactPCI & AdvancedTCA Systems Advanced traffic management aids converged IMS applications

IP Multimedia Subsystem (IMS) board failure, various CoS requirements, supplying the advanced traffic management describes the network infrastructure that and multicast traffic. that converged IMS applications need. supports emerging Internet Protocol (IP) multimedia and telephony applications. In order to meet these goals, advanced RapidIO is designed to be compatible IMS is defined by the 3rd Generation traffic management often makes use of with popular integrated Network Pro- Partnership Project (3GPP) standards the following techniques: cessing Units (NPUs), communications for wireless networks and is also being processors, host processors, and network- applied to wireline networks. n Prioritization and isolation of ing DSPs. RapidIO enables nonblocking traffic based on CoS using queuing switching and guaranteed QoS. Table 1 As applications based on IMS are architectures shows the system requirements for achiev- becoming widely available to consum- n Congestion management using ing guaranteed QoS, and how RapidIO ers, the equipment deployed in the flow control mechanisms that act addresses these requirements. traditional best-effort Internet will need on queues to be upgraded to support the increased n Allocation and management of RapidIO features that enable low latency bandwidth and stringent Quality of resources (bandwidth, buffers) to include: Service (QoS) requirements. optimize system performance n Low overhead protocol resulting in To support IMS, new wireless cell phones Network element systems that implement high fabric utilization and multimedia platforms will integrate advanced traffic management often sup- n Hardware-based data encapsulation combinations of technologies, including: port multiple ports, which are used for with a small Maximum Transmission aggregation or interconnection of vari- Unit (MTU) size for reduced latency n Bluetooth ous networks together in the Internet. A variability n Wireless USB popular standard interface used inside n Hierarchy of flow control features n Ultra Wideband such systems is RapidIO. Defined by the for congestion avoidance and n 2.4 GHz and 5 GHz wireless LANs RapidIO Trade Association, RapidIO is congestion management n GPS location an industry standard fabric interconnect n High-Speed Downlink Packet Access (HSDPA) Requirements How RapidIO addresses requirements n True mobile video broadcasting Low system-level latency Lightweight protocol stack implemented in hardware technologies High link bandwidth to support broadband 1 and 4 lane SerDes-based serial links support 2.5 and applications 10 Gbps data rates. For wireline terminals, high-definition Next generation physical layer Next generation physical layer SerDes with support of up to video, DSL, and Passive Optical Network 80 Gbps data rate per port (PON) enable and drive high-bandwidth Minimum bandwidth guarantee Virtual Channel (VC) architecture consists of scheduling and IMS applications. flow control mechanisms, which result in isolation of traffic from diverse classes of service. Nonblocking architecture Nonblocking system System-level Virtual-output-Queuing (VoQ) and flow control essential enable real-time QoS support With real-time video becoming more Fine granularity traffic management Streams with a few Kbps bandwidth granularity are managed widely deployed, real-time traffic can individually. account for significant network bandwidth, End-to-end flow control and resource End-to-end logical layer flow control between processors which can lead to starvation of data traf- management fic. As such, a combination of strict priority Multicast Transport layer supports up to 64K connections that can be used as unicast connections or to designate multicast groups. and minimum bandwidth guarantees will Switches are capable of multicast. become crucial and the system must be Interworking with other packet interfaces Data streaming encapsulation and physical layer nonblocking to support deterministic QoS. with QoS support (segmentation and reassembly) High availability, redundancy, hot swap, and Fault detection and dual star configuration To realize the required QoS of a mixture of failover traffic with various Class of Service (CoS), In-order packet delivery Supported by physical layer advanced traffic management architectures are often used. The challenge is that QoS Lossless transmission at the link and Supported by physical and logical layers end-to-end levels guarantees must be preserved in the pres- ence of network congestion, component or Table 1

18 / CompactPCI and AdvancedTCA Systems / October 2006 Technology Update

RapidIO offers hierarchical flow control main road. If the car on the side street hub switch’s output port 1 is congested, architecture at the link and logical layers. arrives at the intersection and does not the hub switch conveys VoQ flow control The link layer’s transmitter based flow have the right-of-way (for instance, if the information to the local switch, which control makes pipelining effective. Credit vehicle encounters a stop sign), it must should halt the traffic in the queues cor- based Virtual Channel (VC) and Virtual- wait for a gap in the main street traffic in responding to hub output port 1. The output-Queuing (VoQ) flow control are order to turn left. This wait could indefi- local switch’s multiple queues, one per also supported. The logical layer (proces- nite, which can lead to starvation. hub switch output port, correspond to sor to processor) supports Xon/Xoff of the multiple lanes at an intersection for specific flows, bidirectional arbitration In a communication system, real-time cars going left, straight, and right. The for flows, and end-to-end flow manage- packets, are usually given the highest link between the local switch and the hub ment of classes, streams, and ports. priority in order to guarantee the lowest switch represents the intersection. If any latency. With the resulting strict-priority one of the hub switch’s output ports is The associated flow control mechanisms scheduling, a data packet, must wait until congested, the packets in the other queues of the VC and VoQ architecture operate there are no other higher priority packets of the local switch are not affected and at the link layer. What’s more, they can (such as real-time packets) destined for can still proceed as usual. provide dramatic performance benefits the same output before it can be transmit- at the system level, while adding mini- ted. Thus, data packets can experience Similarly, to support the virtual channel mal overhead. VC architecture allows for starvation, receiving zero bandwidth for a architecture in RapidIO, a separate set of minimum bandwidth reservation for each potentially long time. The Ethernet proto- queues is allocated for each VC. While class of service. VoQ architecture allows col does not specify the scheduling for its RapidIO provides support for up to nine for system-level head-of-line blocking priority queues. PCI Express provides the VCs. In this simple example, there are avoidance to guarantee deterministic option of minimum bandwidth for its VC two VCs. In the local switch, each VC latencies for real-time traffic. queues. Implementations of only strict hierarchically supports a group of VoQs priority for these standard interfaces will for the hub switch’s output ports. VoQ and VC architectures suffer system-level blocking. Consider a one-lane road with cars arriv- The hub switch, which can maintain the ing at an intersection. If the first car in Combination approach same queuing architecture as the local line should turn left, and is blocked from On the other hand, consider that data pack- switch, should schedule packets for each doing so (due to heavy traffic in the left ets will be scheduled to be transmitted at VC according to its minimum required road), then all of the cars behind the first particular intervals. Thus, data traffic will bandwidth. VCs that correspond to data car must wait, even though the other roads not be starved and can be guaranteed a applications are usually the most in need (straight and right) are clear of traffic. minimum amount of bandwidth specified of minimum bandwidth guarantees. Real- by the QoS requirements of the particu- time traffic scheduled according to strict Translating this example to a communi- lar CoS to which the data packets belong. priority could grab much of the system cation system, the cars represent packets, The RapidIO VC architecture allows for bandwidth and potentially starve data the intersection represents a hub switch, the system to guarantee minimum band- packets indefinitely. However, the sched- and the backed-up line represents a First- widths, while also preserving the low uler must also make sure that interactive in-First-out (FIFO) queue. Real-time latencies required for real-time packets. and streaming packets are transmitted packets that are blocked in the FIFO To do this, often a combination of strict with the required levels of low latency. would suffer latencies, which are nonde- and minimum bandwidth scheduling will Erlang has implemented schedulers that terministic. The packets would likely fail be employed VC and VoQ architectures support the requirements noted earlier as to achieve their required levels of QoS. can be implemented in a communication a combination of strict priority and mini- Examples of standard protocols, which system to improve system performance. mum bandwidth scheduling. suffer from system-level blocking, are Ethernet and PCI Express. A communication system, typically If the hub switch experiences excessive consists of several line cards. The line congestion for a particular VC, flow con- On the other hand, consider a road with cards communicate to each other as trolling that VC lets traffic belonging to multiple lanes. Cars arriving at the inter- peers through a shared hub switch. Each the other VCs traverse the system without section go to the lane corresponding to line card may have a single processor or violating their QoS requirements. Con- the turn they plan to make. If the car turn- multiple processors used as a farm. The trol takes place by having the hub switch ing left is unable to do so because of con- processors are usually connected to a bridge convey VC flow control information to gestion, the cars in the other lanes are still or a local switch on the line card, which is the local switch for a particular VC. The able to go because they have another path in turn connected to the hub switch over a local switch’s scheduler will then react to to their destination using the other lanes. RapidIO backplane. The interface between the flow control by limiting the number Implementing the RapidIO VoQ architec- the processor and the bridge and between of packets transmitted for that VC. This ture enables a nonblocking system. the processors and the local switch can be can only be done if the local switch has RapidIO as well in order to reduce system queued packets for each VC separately. If To understand the advantages of the cost and improve performance. the hub switch experiences congestion for Virtual Channel architecture consider VC1, it sends flow control information to a road with a continuous stream of cars In order to support VoQ, the local switch the local switch, which temporarily stops passing through an intersection that have allocates a separate queue per hub switch transmission of all packets belonging to the right-of-way because they are on the output port. In this simple example, if the VC1 queues.

CompactPCI and AdvancedTCA Systems / October 2006 / 19 Technology Update

To model IMS systems with a mixture tion, the latency is about 100 microsec- requires more start-up time in the pipe- of voice, video, and data traffic, it is onds at 55 percent load and quickly goes line. However, the latency will be lower important to simulate bursty traffic. Data to infinity thereafter. than the base case at loads greater than transfers and variable bit-rate video traf- 25 percent and can reach 80 percent or fic tend to be bursty in nature. The simu- A switch architecture that supports VC higher loads. With additional scheduling lation assumes an average burst length flow allows for improved performance, and bandwidth management optimiza- of 20 packets being transferred between with about 100 microseconds latency at tions, Erlang has been able to achieve line cards. If each RapidIO packet size is 65 percent load. This is an incremental 95 percent and greater loads with less 256 bytes for the payload, this would cor- performance improvement, however, the than 100 microseconds latency in high- respond to 20 × 256 = 5,120 bytes being VC architecture allows for minimum performance switch fabrics. transferred, which could be the size of bandwidth guarantees for traffic that four video packets (each of 1,280 bytes), is not real-time (data), while preserv- Peter Yan is an active member of the transmitted in a burst. ing the QoS for real-time traffic (voice RapidIO Trade Association and chief and video). This can be achieved with a technology officer at Erlang Technology. For the base case, consider a system combination of strict, plus Deficit Round architecture that does not support any Robin (DDR) or Weighted Round Robin To learn more, contact Peter at: flow control. The system implementa- (WRR) scheduling, which Erlang has tion could be simpler and may be able implemented in switch fabrics and traffic Erlang Technology to achieve the lowest latency at very low managers. 345 Marshall Avenue, Suite 300 loads. For instance, at 1 percent load, it Saint Louis, MO, 63119 was assumed that this switch architecture Finally, the switch architecture, which Tel: 314-276-5583 has a fall through latency of 200 ns. How- supports VC and VoQ flow control, Fax: 314-336-5902 ever, with increased levels of congestion starts at 800 ns latency at 1 percent E-mail: [email protected] and the inability to control the conges- load because the scheduling algorithm Website: www.erlangtech.com

RSC# 20 @ www.compactpci-systems.com/rsc

20 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 21 @ www.compactpci-systems.com/rsc RSC# 21 @ www.compactpci-systems.com/rsc GUEST

TELECOM SOFTWARE How DSO, COTS, and open architectures can help solve the looming telecom software crunch

By Michael Christofferson

Complex distributed telecom and engineering staffs, many equipment An open architecture platform’s layer networking applications require a new makers are embracing a new development of abstraction clearly separates telecom method of development. Michael makes process. An approach known as Device applications and the underlying hard- the case that by using industry standard Software Optimization or DSO, repre- ware and system software. This layer interfaces and open architectures, NEPs sents a fundamental rethinking of the of abstraction, coupled with the use of and TEMs can greatly accelerate the design process, leveraging products and standard interfaces, enables designers to development of applications and systems. practices that embrace reusable code, open use best-of-breed COTS hardware and standards, and preintegrated Commercial software from multiple vendors. It also Worldwide, most network equipment Off-the-Shelf (COTS) technology, stan- enhances portability, allowing NEPs to makers have already abandoned the prac- dardized across the enterprise (not just upgrade hardware and software at later tice of creating their own proprietary across a single development group). dates with minimal disruption to their DSPs, network processors, operating sys- proprietary applications. tems, protocol stacks, and management Equipment makers using a combined DSO tools. Many, however, are still creating and COTS paradigm can trim their plat- Open architecture telecom platforms significant pieces of their application- form teams by up to 80 percent. Manpower provide the basic components needed to specific hardware and software infra- cost is reduced, and NEPs can allocate develop and host telecom applications structure from the ground up, constantly their engineering resources to value-added and services. Figure 1 shows the compo- inventing and reinventing the wheel application and service development. nents typically represented by the Enea from one project to the next. For these Networking Application Services Plat- equipment makers, many of whom are Advantages of an open form (NASP) platform: trimming engineering staffs in order to architecture platform for telecom contain costs, this homegrown approach By using a preintegrated DSO telecom n Operating systems is becoming increasingly untenable. platform, telecom developers can shorten n Interprocess Communications (IPC), the lengthy platform design/integration such as Enea LINX and Open Source According to industry analyst Venture process into an evaluation and purchasing IPC technology Development Corporation (VDC), the process that can take as little as two n High Availability (HA) middleware amount of code deployed in today’s net- months. All told, this COTS approach can framework that is Service Availability work equipment is growing exponentially – shorten the application development cycle Forum (SAF) compliant currently accounting for more than half of by up to 50 percent. n Database management system total project costs. Meanwhile, the number of available developers is growing at a relatively flat pace, thereby creating a serious mismatch that makes it increasingly difficult to meet time-to-market windows. Already, VDC estimates that 50 percent of device projects fall behind schedule by an average of four months.

The world’s approximately 600,000 developers cannot keep up with the rocketing demand for device software required by today’s smart cell phones, automobiles, appliances, and entertainment systems. In addition, manufacturers are facing the twin challenge of ever-shortening production cycles to meet competitive marketing pressures while at the same time having to handle exploding device complexity as end users demand new features and sophisticated capabilities. Hardware Platforms – AdvancedTCA, Reference boards To ease this software development challenge, and to make do with leaner Figure 1

22 / CompactPCI and AdvancedTCA Systems / October 2006 Operating systems By using direct message passing, an Ideally, a telecom software platform interprocess communications frame- should support multiple operating systems work’s can increase performance by in both homogeneous configurations, such enabling application processes to com- as those with Linux or an RTOS deployed municate directly with each other on a throughout the system, and heterogeneous peer-to-peer basis, without having to configurations, which combine Linux and synchronize through intermediate mech- one or more RTOSs deployed on multiple anisms such as mailboxes, semaphores/ CPUs, shelves, and blades. mutexes, event flags, UNIX-style signals, or even sockets. This direct approach also This flexibility enables developers to simplifies communications and facili- select the OS or processor combina- tates logical process separation, thereby tion best suited to their application. For enhancing reliability and simplifying example, some NEPs may prefer to run fault recovery, particularly in distributed Linux on one set of blades to host IT- systems utilizing multicore devices and oriented supervisory and enterprise man- complex network topologies. agement functions, while using an RTOS on another set of blades to host DSP- In order to maximize scalability and based media processing applications with portability, the interprocess communi- tight size and performance constraints. cations services should provide transparency, independent of the underlying Interprocess communications processor, operating system, or inter- In a distributed network, an interprocess connect. This transparency enables communications framework like Enea’s distributed platform components and open source LINX can provide the glue applications to communicate in a seam- needed to integrate platform components less fashion, as if they were residing on a and applications across multiple pro- single processor under a single operating cessors, operating systems, blades, and system. When combined with the ability shelves (Figure 2). Ideally, this frame- to dynamically discover communication work should provide dependable, high- endpoints, this transparency also enables speed transport for both the control and developers to locate applications on data plane over reliable as well as unreli- any node in the system, and change the able interconnects and protocols. It should configuration at run time. The result also support the encapsulation of other is that developers and service provid- bearer protocols – such as TCP, UDP, and ers can dynamically change and scale SCTP – for data transport. Because it is the system configuration, redistribute open source, Enea’s LINX also makes it applications across multiple blades, possible to use the IPC framework for a and upgrade the hardware with minimal wide variety of applications and systems. application code changes.

LINX Architecture API Application Flow Control LINX Client Application Payload Byte Order Layer Conversion

Traffic Interfaces Management Interfaces

Naming Link Session Connection Management Service Management Layer Address Address Publication Supervision Supervision Resolution Subscription

LINX Transport Protocol

Transport(s) Fragmentation Fragmentation System Byte Order Layer Conversion Bundling Bundling Encryption Sequence/Retransmission

Unreliable Interconnects Reliable Interconnects Link or Protocols or Protocols Layer SCTP TCP Shared Memory/PCI UDP SCTP Ethernet RIO

Figure 2 RSC# 23 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 23 RSC# 24 @ www.compactpci-systems.com/rsc GUEST

TELECOM SOFTWARE

Middleware framework added applications and service develop- To learn more, contact Michael at: The flexibility provided by the network ment, and get their completed designs to platform’s IPC layer lays the foundation market on time and on budget. Enea Embedded Technology for more advanced distributed commu- c 2635 North First Street, Suite 118 nications, instrumentation, monitoring, Michael San Jose, CA 95134 and high availability services, collectively Christofferson is Tel: 408-383-9480 referred to as middleware. The middle- director of product Fax: 408-383-9485 ware extends the IPC’s process-to-process management, E-mail: [email protected] communications services, providing one- Enea Embedded Website: www.enea.com to-many communications that facilitate Technology. system wide communications, instrumentation, and event notification. These extended communications services, in turn, lay the groundwork for high availability monitoring, detection, recovery, and reporting services that are essential for building a true nonstop computing platform.

To simplify configuration and management at the slot, blade, and chassis level, COTS middleware solutions provide shelf management services, typically utilizing standard interfaces, including SAF’s Hardware Platform Interface (HPI). Complementing the middleware’s shelf management services are heuristic fault management services, which provide monitoring, detection, recovery (for example, restarting a failed application or failing over to a redundant blade), and reporting for every resource in the system.

Database management The network platform may simplify information (data) management and sharing across multiple nodes with a Relational Database Management System (RDBMS). Unlike traditional desktop databases, these embedded databases must often work in a diskless environment and deliver higher levels of performance and availability. Enea’s small-footprint Polyhedra RDBMS, for example, uses a memory- resident design that boosts performance by 10x relative to conventional disk- and flash-based RDBMSs.

A new way of getting the job done In the past, NEPs have been content to create their own OS, middleware, and database solutions. However, with the new challenges of reduced production cycles and ever-increasing software complexity, a growing number have recognized that they can no longer remain competitive by creating, maintaining, and porting platform software in-house. Field-proven, preintegrated COTS network platforms allow NEPs to outsource their platform design, focus engineering resources on value- RSC# 25 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 25 SPECIAL

MEDIA SERVERS AND SOFTSWITCHES I-TDM: Supporting TDM voice in the age of MicroTCA and AdvancedTCA

By Robbie Dhillon, Ian MacMillan, and Amir Zmora

The ecosystem that is growing voice and video traffic from TDM to IP Media resource functions around the I-TDM standard now networks, to connect between the follow- Media resource functions perform IMS includes both hardware- and software- ing networks: network media services, such as voice based implementations. As with any mail and recorded announcements. Many new standard, initial implementation n Broadband wireline IP network, MRFs are designed to support both IP and work involves identifying areas where mainly using SIP but also using TDM interfaces, although implementing the standard needs enhancement. legacy H.323 equipment an MRF with IP support only is possible. The authors outline the role PICMG n Broadband wireless IP network SFP.1 – also known as Internal Time (Wi-Fi and WiMAX), typically Wireless radio network controllers/ Division Multiplexed circuit switching using SIP base stations – plays in converting voice and media n 3G real-time conversational voice Wireless radio network controllers manage packets for transport and storage. and video, using 3G-324M currently and connect with wireless base stations. and SIP in the future While the base stations will migrate to IP Demand for voice and video over IP n Voice PSTN network transport over time, many existing base along with the adoption of IMS archi- – Access media gateways that have stations use TDM transport to connect tecture for next-generation networks TDM telephone lines or TDM trunk with the RNC. Base stations performing are driving the trend towards network interfaces IP transport can use I-TDM to transport components based on standard form fac- – Border media gateways that voice internally to the base station. tors. Many telecom equipment manufac- interconnect TDM networks to turers who have traditionally designed IMS networks. For example, Conferencing servers proprietary platforms internally are now connecting a TDM wireline network Similar to MRFs, many conferencing seeking off-the-shelf components based to an IMS-based wireless network servers are designed to support both IP on standard telecom grade, high capacity – Mobile video media gateways and TDM interfaces. form factors. As a result, the market is that translate voice and video adopting MicroTCA and AdvancedTCA bidirectional and streaming traffic I-TDM allows these IMS platforms to platforms, with large deployment numbers between IP (typically SIP) networks be implemented using MicroTCA or predicted for 2007 onward. and mobile 3G-324M networks. AdvancedTCA by taking the TDM voice The 3G-324M network has a fixed and media and converting them into A number of key IMS architecture ele- limited bandwidth of 64 Kb and packets for transport and storage within ments handle voice and media traffic. screen size is typically limited to the platform. In order to interconnect with voice and QCIF resolution, so the gateway media traffic from TDM networks, the must perform two key tasks: I-TDM building blocks IMS architecture provides platforms that a. Transcoding: Voice to The IMS platforms we have described interface with these TDM networks and NB-AMC and video to incorporate two types of building blocks support voice, media, and signaling. H.263 or MPEG-4 that must support I-TDM: b. Modification: Changing the MicroTCA and AdvancedTCA are packet- frame rate, typically to 10 FPS n I/O interfaces, which provide the based, causing IMS systems to transmit and the resolution to QCIF physical interface to TDM networks TDM via I-TDM running over packet trans- n Video over PSTN using H.324 n Digital signal processors, port, such as GbE. There is now a growing is deployed in countries where which transform the TDM media ecosystem around the I-TDM standard broadband penetration is limited for storage and transport in IMS with the availability of both hardware and networks software-based implementations. In the case of video over PSTN, deployed in Italy and the UK, the H.324 streams I/O interfaces IMS voice and media platforms are sent using modem connectivity over In MicroTCA and AdvancedTCA plat- There are a number of IMS platforms PSTN lines. As in 3G-324M control, forms, I/O interfaces are typically imple- that interface with TDM voice and voice and video are multiplexed and sent mented as Advanced Mezzanine Cards media. over the network. Connecting this com- (AdvancedMCs). The AdvancedMCs munication with other networks such as take the TDM interfaces and convert the Media gateways broadband SIP requires gateway func- TDM voice to I-TDM for transport over IMS and other NGN networks require a tionality similar to that used by the mobile the backplane of the MicroTCA and media gateway, which typically translates video media gateway. AdvancedTCA platforms.

26 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 27 @ www.compactpci-systems.com/rsc SPECIAL

MEDIA SERVERS AND SOFTSWITCHES

Bundling signal channels with the voice intensive applications such as media n External memory per DSP on board, bearer channels enables AdvancedMC gateways and media servers. The DSP enabling easy addition of features interface cards to add significant value to farm handles all media processing func- without concerns about code and the MicroTCA/AdvancedTCA platform tionalities that enable convergence of data size by providing intelligent protocol off- IP (wireline and wireless), PSTN, and n A DSP software framework that load and acceleration. The AdvancedMC mobile networks. This board must per- enables: processor and protocol software must form not only simple transcoding of voice – Adding user-defined and proprietary be sophisticated enough to handle the and video, but also: algorithms to the DSP TDM network’s large number of differ- – Support for predictive scheduling, ent protocols. n Voice event detection and generation, allowing tasks such as DMA of DTMF detection, CNG, VAD, and relevant data to be sequenced The typical standard TDM interfaces ECAN automatically while a previous task include T1/E1/J1 and OC-3/STM-1 or n Video frame rate and resolution is being processed OC-12/STM-4. change, text overlay, alpha blending, – Support for real-time processing cropping, picture-in-picture with guaranteed quality of service The Interphase iSPAN 3639 AdvancedMC n Media server tasks such as streaming, and latency Multiprotocol T1/E1/J1 Intelligent Com- recording, and conferencing of voice munications Controller, available in and video quad or octal configurations, (Figure 1) n Fax and modem processing is an example of an intelligent protocol offload/acceleration AdvancedMC. This In MicroTCA/AdvancedTCA platforms, card handles both: the media processing capabilities are usually implemented on AdvancedMC n Signaling protocols such as SS7 and cards populated with DSPs. The SurfRider ISDN with an onboard processor AdvancedMC from Surf Communication n I-TDM for up to 256 DS-0 flows with Solutions (Figure 3) for example can sup- Figure 3 an onboard FPGA port the convergence of PSTN, IP, and wireless networks, and must interface to the IP and TDM worlds by supporting I-TDM standard brief I-TDM. I-TDM enables the passing of the The legacy solution for carrying TDM TDM traffic over the AdvancedMC card’s on the backplane has been the H.110 bus, packet-based interfaces. Dedicated mech- with 32 TDM data highways at 8 MHz. anisms on the AdvancedMC cards convert Carrying TDM over a serial packet bus the incoming I-TDM packets to TDM reduces pin count, increases reliability, Figure 1 traffic, and vice versa. and allows the prospect of a single fabric for carrying all chassis traffic. The advent Wireless networks introduce an addi- An AdvancedMC DSP board for media of MicroTCA and AdvancedTCA has tional complication by using subrate processing functions must support: increased the urgency of the problem, DS-0 TDM voice. AdvancedMCs that can because unlike CompactPCI, there is provide subrate TDM voice interfaces n A flexible number of channels, no defined TDM highway backplane in address this issue and have enhanced the I-TDM flows, and different mixes of AdvancedTCA/AdvancedMC. I-TDM protocol to transport the subrate 1 ms and 125 µs flows voice. For example the Interphase iSPAN n A variety of external interfaces: PICMG chose a layered, fabric-agnostic 3632 Quad-Port Channelized OC-3/ Some customers may require the approach to standardize a TDM-over- STM-1 Interface Processor (Figure 2) TDM to be moved over Ethernet packet protocol. The lower layer speci- provides up to 8,032 I-TDM flows. packets; others may require different fication, SFP.0, defines essential fabric packetization support (such as SRIO, services, such as detection of misrouted InfiniBand) or dropped packets, multiplexing, and n The ability to dynamically and end-to-end fabric integrity check. PICMG simultaneously run all media types deliberately chose a lightweight proto- to allow moving from 100 percent col to run directly on top of the Layer 2 voice to 100 percent video, and any packet. For example (Figure 4), SFP.0 mix in between runs directly over MPLS over Ethernet, n Separation between media and control without requiring a bulky IP or TCP/UDP functions and utilizes DSPs with a layer. Figure 2 direct IP network interface to avoid aggregation on the host processor, The next protocol layer, SFP.1, also known Digital signal processors which typically creates bottlenecks in as I-TDM, provides TDM-centric header A digital signal processor farm is an the system and increases media delay, services. I-TDM multiplexes several essential component in media processing- reducing quality TDM channels together into one packet,

28 / CompactPCI and AdvancedTCA Systems / October 2006 SPECIAL

MEDIA SERVERS AND SOFTSWITCHES

32 16 0-32 8 1 1 1 2 11 16 24

Layer 2 MPLS Time- Sequence Last Packet Check Pad SOP EOP Reserved UID Payload Header Label Stamp Number Packet Length Sum

Layer 2 SFP Conveyance SFP Conveyance SFP Payload Headers Header 1 Header 2 NPF Header (specified by UID) (10 Bytes) (5 Bytes) (3 Bytes) Figure 4 as opposed to waiting several frame times backplane technology will allow system changes and extensions inherent in a to accumulate enough data for a given vendors to use interoperable vendor parts newly introduced standard. TDM channel. This step maintains fabric and concentrate on end functionality over efficiency, because most fabrics (and net- plumbing. The NP-based approach is ideal for work hardware such as switches and end- supporting a mix of 1 ms and low density points) become increasingly inefficient I-TDM enabling technologies 125 µs modes of operation. The FPGA- with small packet sizes. Figure 5 shows Early implementation of the I-TDM based method scales better to higher density an I-TDM 125 µs mode payload format. standards has included software-only 125 µs implementations. I-TDM payload format choices are: implementations using dedicated network processors and reprogrammable FPGAs A software implementation of 125 µs n A packet emission interval of 125 µs (Altera and Xilinx). Both approaches I-TDM in Wintegra’s WinPath NP family provides the lowest possible latency, have the flexibility to accommodate the (Figure 6) supports densities of up to and is recommended for hardware implementation 8 1 1 1 1 1 1 1 1 1 1 1 1 1 (Bytes) n Packet emission interval of 1 ms, I-TDM Channel TDM TDM TDM TDM TDM TDM TDM TDM TDM TDM TDM TDM TDM which creates higher latency, but L2 is easier to handle in host media Headers Management Byte Byte Byte Byte Byte Byte Byte Byte Byte Byte Byte Byte Byte processing and other software-centric approaches Up to 512 TDM Bytes in an I-TDM 125 µs Mode Packet Unlike WAN-centric TDM over packet standards such as PWE3/TDMoIP, (Bits) I-TDM is intended as an in-chassis pro- Flow ID 2 1 1 4 24 7 9 7 9 Reser tocol. Therefore, there is no support for Specifies Activ

ACK Channel Destination Channel Channel Reserved Channel ate timing recovery, which greatly simplifies ved Reserved Location 2/ Jitter Queue Command Location 1 the implementation. ID Size

Command Set The I-TDM standard primarily benefits 0 = Reserved designers as a chassis-optimized, fabric- 1 = New Channel_ID at Byte Offset Channel _Location_1 with Size = 1 neutral technology supported by mul- 2 = Close Channel_ID at Byte Offset Channel_Location_1 with Size = 1 tiple vendors. It is possible to transport 3 = Relocate Channel_ID from Byte Offset Channel_Location_2 to Channel_Location_1 TDM over packet in multiple nonstan- 4 = Cyclic Reaffirmation: Channel_ID is at Byte Offset Channel_Location_1 with Size = 1 dard ways, and point solutions have been 5 = Packet Rate Integrity Check 6-15 = Reserved implemented in the industry. Widespread adoption of I-TDM as the TDM transport Figure 5

Memories

1 x OC12/STM-4, R

SBI EMPHY2 R Backplane Gigabit Ethernet 4 x OC3/STM-1 or Framer Interface TM to Backplane 12 x DS3/E3 Mapper WIN867M6

UFE3

Figure 6

CompactPCI and AdvancedTCA Systems / October 2006 / 29 SPECIAL

MEDIA SERVERS AND SOFTSWITCHES

28 T1/21 E1s (672 DS0s) with DS0 granularity per I-TDM channel and up to 84 T1/63 E1 with 8xDS0 granularity per TDM TDM Microsequencer I-TDM channel. Wintegra also supports Highway Slip Buffer/ higher densities of up to OC-12 (8K (H.110) Timeslot PKT Jitter Buffer Dual Generic channels) by using an I-TDM load in its Memory I/F Packet Bus: UFE FPGA. (8-bit, 125 MHz) Low Pin-count Figure 7 shows Accolade Technology’s Interfaces FPGA-based ASTDM solution. The ASTDM features dual GbE interfaces To-Switch Parser to the fabric switch side, a TDM bus From-Switch Pkt (signaling termination, DSP resources, or RAM Timeslot Mem SONET framers), and a CPU interface. Memory This architecture uses dedicated state machines to manage both 1 ms and 125 µs modes of operation while directing command and control packets to the CPU interface. It can support densities of four T1/E1s to OC-12. CPU I/F DMA State of the I-TDM ecosystem The current I-TDM ecosystem includes manufacturers of enabling technologies and board level products. Suppliers of IP cores for FPGAs and software imple- Generic 32-bit mentation on network processors that CPU Interface have announced I-TDM solutions include Accolade Technology and Wintegra Inc. Figure 7

Suppliers of TDM, signaling interfaces, data rate. However wireless networks Summary and media server/DSP blades that have widely use subrate channels (fractional Platforms designed around MicroTCA announced AdvancedTCA/AdvancedMC DS0, such as 32 Kbps, 16 Kbps, 8 Kbps). and AdvancedTCA are rapidly coming to board level solutions with I-TDM func- Currently, the I-TDM standard cannot the market. I-TDM provides a key tech- tionality include Interphase Corpora- specify channel ID or carry out 125 µs nology to enable IMS next-generation tion and Surf Communications, among mode channel management commands network platforms based on MicroTCA others. like New, Change, Relocate, or Cyclic and AdvancedTCA to successfully lever- Reaffirmation on a sub-DS0 granular- age the significant investment in legacy TDM interface vendors and DSP ven- ity. While it is still possible to trans- TDM voice and media networks, while dors are testing I-TDM interoperability. port sub-DS0 channels by some private at the same time being able to provide Accolade Technology is introducing an understanding between the two peers, a the advanced services and flexibility I-TDM reference test platform based standardized scheme promotes interop- inherent in the MicroTCA/AdvancedTCA on its ASTDM implementation of the erability and possibly reduces wasted architectures. I-TDM standard on a Xilinx-based bandwidth. development platform. This menu-driven A growing number of critical IMS I/O I-TDM reference test platform supports Redundancy and protection switching and DSP building blocks are incorporat- call setup and I-TDM traffic generation support ing I-TDM solutions. and termination. While specific redundancy schemes at the chassis level are not precluded by the Since the standards and the building Future enhancements to standard I-TDM standard, they are not explicitly blocks for these solutions are currently The current I-TDM standard is suffi- supported. in the initial stages of market penetration, ciently mature for product implementa- innovative and flexible implementations tion. However, for optimal functionality Modifications or extensions to will pave the way for additional capa- and interoperability, additional issues the 125 µs channel management bilities in the near future. As with any should be considered. command scheme new technology, the I-TDM standards I-TDM’s 125 µs channel management will likely continue to evolve based on Explicit support for subrate channels in-band nature is a strong architectural implementation experience in voice and The most common channel format is a benefit. However, improvements in some media transport within MicroTCA and complete DS0, equivalent to a 64 Kbps modes of operation may be desirable. AdvancedTCA platforms. I-TDM clearly

30 / CompactPCI and AdvancedTCA Systems / October 2006 SPECIAL

MEDIA SERVERS AND SOFTSWITCHES

I-TDM ECOSYSTEM GLOSSARY Amir Zmora is BSC Base Station Controller Surf Communication Solutions VP CNG Comfort Noise Generation business develop- DMA Dynamic Memory Allocation ment, responsible for all business DSP Digital Signal Processor development, DTMF Dual Tone Multi-Frequency including estab- ECAN Echo Cancellation lishing partnerships with telecom infrastructure hardware manufacturers. FPGA Field Programmable Gate Array Amir is a well-known speaker at VoIP GbE Gigabit Ethernet events, and has published articles in GGSN Gateway GPRS Serving/Support Node numerous industry magazines. Amir holds an MBA from Tel Aviv University, IMS IP Multimedia Subsystems a BA in Economics from Hebrew IP Internet Protocol University, and a BA in Software I-TDM Internal Time Division Multiplexed Engineering from the Sela Institute. MRF Media Resource Function To learn more, contact the authors at: NP Network Processor Robbie Dhillon PSTN Public Switched Telephone Network Accolade Technology PWE3/TDMoIP Pseudo Wire Emulation Edge to Edge/Time Division Multiplexed traffic over IP 8 Woodfield Road QCIF Quarter Common Intermediate Format Acton, MA 01720 Tel: 508-479-9663 RNC Radio Network Controller E-mail: [email protected] SIP Session Initiation Protocol Website: www.accoladetechnology.com SONET Synchronous Optical Networking Ian MacMillan SRIO Serial RapidIO Interphase Corporation TCP/UDP Transmission Control Protocol/User Datagram Protocol 2901 North Dallas Pkwy, Suite 200 Plano, TX 75093 TDM Time Division Multiplexed Tel: 214-654-5000 TEM Telecom Equipment Manufacturer E-mail: [email protected] VAD Voice Activity Detection Website: www.iphase.com VoIP Voice over Internet Protocol Amir Zmora Surf Communication Solutions. offers the benefits of a cost-effective, Ian MacMillan 496 Old Connecticut Path, Suite 320 standards-based solution for offering new is the senior Framingham, MA 01701 services while protecting investments in product marketing Tel: 866-644-3379 today’s infrastructure. manager at E-mail: [email protected] c Interphase Website: www.surf-com.com Robbie Dhillon Corporation is VP of business responsible for the www development AdvancedTCA/ at Accolade AdvancedMC and Signaling Solutions Hot links at Technology, product lines. Ian’s broad telecommuni- advancedtca-systems.com with more than cations background spans more than 20 years of 15 years and includes work for large Legacy telecom hits the telecommunications TEMs and carriers, including Nortel 21st century: TDM circuits on and networking experience. Prior Networks and Verizon. Ian has worked AdvancedTCA switched fabrics to cofounding Accolade, he was with emerging technologies covering all president of Network Modules Inc. sectors of telecommunications including advancedtca-systems.com/ and VP of business development at voice, data, and wireless. articles/id/?374 SBS Technologies. Robbie has a BSEE and MS in Physics from Portsmouth University, UK. SPECIAL

MEDIA SERVERS AND SOFTSWITCHES Selecting a modular media gateway to enable VoIP and other content-rich media services

By Venkataraman “VP” Prasannan

Here Venkataraman argues the case tocol (IP) networks enable voice, data, effective open systems standards such as for a modular media gateway architec- wireless, and increasingly, video. AdvancedTCA. ture that can help developers drive down costs for VoIP and support other When considering VoIP for reducing At the core of media gateways is an enor- content-rich communications services. communications costs, companies need mous amount of signal processing. This to consider these questions: is performed by Digital Signal Processors With continued globalization, many cor- (DSPs) to provide the traffic interface porations have workgroups dispersed n Is the media gateway under between the Internet and the wireless net- around the world. To support work- consideration a comprehensive and work of the PSTN, and to convert Time- group productivity and communications, cost-effective solution with a low Domain Media (TDM) voice to VoIP and companies are considering simpler, less price but no price penalty for a fully back. This process is time-dependent and expensive network solutions that com- loaded box? the system must re-assemble packets in bine content-rich voice, wireless, data, n What is the existing voice and data their time-stamped sequence (Figure 2). and video. traffic on our networks? When packets are missing or late, the n What are the current voice and data media gateway discards and interpolates These businesses expect the services they usage patterns? them. The media compresses voice pack- select and implement to have the high n Will the quality of the VoIP audio be ets, corrects for jitter, cancels echoes, and quality of the telephone and other ser- acceptable? generates or recognizes tones. vices they now use, and they expect them n Are the reliability and uptime high to boost productivity and help reduce enough to meet our requirements? Packetizing such time-dependent media overall communication costs. as voice, wireless, and video data is more Media gateways bridge multiple complex and requires special processes, To access the desired network services technologies such as time stamping. In short, the order requires connections from wireless, A media gateway provides the conversion in which the time-dependent media IP wireline networks, and the Internet to and switching of voice media between a packets arrive determines their meaning. the Public Switched Telephone Network network and its access points. For those Without proper sequencing, these pack- (PSTN) with flexible, robust, scalable, using computers, Digital Subscriber ets become unintelligible gibberish, not and cost-effective media gateways. These Line (DSL), or cable modems, a media unlike a sentence made up of randomly gateways must reduce communication gateway converts, compresses, and pack- generated phonemes that ignore linguistic cost, increase the effectiveness of the etizes voice data for transfer back and rules. All ingress and egress data type and dispersed workgroups, and form the forth across the Internet backbone for formats – including TDM from the RAM, foundation for future media services. wireline or wireless phones. Media gate- Internal Time-Division Multiplexing ways fit between the PSTN and wireless across the backplane, TDM, or (AAL2) Juniper Research estimates that the or IP-based networks (Figure 1). Because and Real-time Transport Protocol (RTP) business VoIP market will grow from of the need to reduce costs, engineers are – have time-stamp information associated $4.5 billion in 2004 to $20 billion by today building media gateways on cost- with each packet and/or cell. 2009. What’s driving this growth? The cost of VoIP service is dropping, and its quality of service is continuously improv- PSTN ing. What’s more, VoIP costs less to run, Wireline RAN install, and scale. Attracted by their near Wireless wireline quality and lower cost, businesses AIN-SS7 are already considering or employing PABX offerings from vendors such as Vonage, DeltaTree, Net2Phone, and 8x8.

Initial VoIP considerations Internet Enterprise Today, all communication can take place (PVT VoIP Network) through a computer. Widely accessible broadband technology and Internet Pro- Figure 1

32 / CompactPCI and AdvancedTCA Systems / October 2006 MEDIA GATEWAY DATA PLANE

iTDM iTDM TDM TDM or AAL2/ATM RAN STM-1 STIM-1 DSP (Wireless) Card Card

Media Gateway CPU CPU Controller

MEDIA GATEWAY PSTN Media Gateway CPU CPU Controller

iTDM iTDM RTP

INTERNET IP DSP STM-1 NPU Card TDM or AA:2/ATM

Figure 2

Why modular media gateways n What is the expected return on for VoIP? investment from VoIP? Several market demands are pushing n Have we identified the total cost of companies to converge all of their media ownership including soft cost savings services using media gateways, beginning from easier system expansion and with VoIP. Companies are expecting this updates and reduced cost of space? architecture to: Investing in the future n Lower initial costs. Capital investment In the past, network applications dedicated is less because hardware can be used separate, often proprietary and costly for multiple functions. networks for each application. Modular n Lower development costs. media gateways’ horizontal organization Open system hardware and software offers the advantage of a common archi- standards with well-defined tecture for all applications. Today, media application mean lower costs, and gateways can ensure the inter-working of Application Programmable Interfaces transactions between different technolo- (APIs) accelerate development. gies by enabling services for media-based n Handle multiple media types. applications involving voice, video, fax, Companies want VoIP solutions today, and data. but need to select solutions that will also handle video in the near future. As a bridge between the PSTN and wire- n Lower the costs of deployment and less networks, the media gateway connects maintenance. Standardized, modular to the Radio Network Center (RNC), the systems reduce training costs and Mobile Switching Center (MSC), and the maintenance while also improving PSTN. Connecting the PSTN and Radio uptime. Access Network (RAN) to the Internet n Enable faster time-to-market. Early requires vendors to design and market market entry hits the window of less costly, modular, and configurable opportunity and maximizes revenue. media gateways that handle more voice and video. VoIP business issues The process of selecting a VoIP media To meet the current and future dollar-per- gateway platform begins with developers port value needs, media gateway designs answering several business questions, should include multicore DSPs and power- including: ful Network Processor Units (NPUs) to distribute data, handle multiple data- n What are the business drivers for transport protocols, increase the number implementing and providing VoIP? of data ports, and handle more bandwidth. n Have we determined the immediate To protect a buyer’s investment, this also and future data capabilities we will requires developing media gateways need? that are flexible enough to scale up from n What are the security concerns and today’s 100 megabit networks to 1 giga- challenges for our business? bit, and even multiple gigabit networks. RSC# 33 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 33 SPECIAL

MEDIA SERVERS AND SOFTSWITCHES

Considerations for modular VoIP volume may influence your architecture Summary media gateways choice, your ability to scale up or down Today, companies of every size are under Companies need to answer several ques- by adding new combinations of cards and pressure to manage costs while increas- tions when considering a media gateway chassis, as well as the system cost. ing workgroup and team interaction to solution. improve productivity. The fundamental An implementation of a modular factors driving the shift to VoIP and other What solutions are available for the media gateway media services are lower capital costs, application? One example of an application-ready reduced operational costs, and improved A carrier-grade shelf-and-server platform modular media gateway that processes business processes, especially a com- meeting 5-nines (99.999 percent) per- 1,000 channels or more per blade is the pany’s internal and external communi- formance, and offers expansion capacity RadiSys ATCA-1200 AMC Carrier blade. cations and collaboration. Integration to add more processors and edge cards. Based on today’s STM-1 card (155.52 mbps of media services into the realm of IT is AdvancedTCA and MicroTCA architectures capacity) and DSP technology, this design now possible through the use of media offer this. A combination of AdvancedMCs scales up or down easily. In addition to gateways. in an AdvancedTCA or MicroTCA chassis VoIP, these gateways support data, fax, and is also a possible solution. video moving through the network, includ- A standards-based media gateway pro- ing managing TDB sequencing. vides companies with a platform on What is most cost-effective for the which to build the converged, content- application? RadiSys’ typical modular media gate- rich media applications they need – This depends on the application’s call way architecture includes components including VoIP – while gaining cost volume and the need to scale it up or down. to address flexibility, scalability, and savings, quality, and scalability. AdvancedTCA handles volumes of more processing needs. c than 10,000 calls. MicroTCA will easily Venkataraman handle call volumes of up to 1,000 calls. Figure 3 depicts a media gateway “VP” Prasannan The advantage of AdvancedTCA is the high implemented with four AdvancedTCA is the senior density it enables – 8,000 calls or more per AdvancedMC carrier blades and the director of Product slot. However, the same advantage could appropriate AdvancedMC cards populat- Line Management become a drawback if the requirement is ing them. The AdvancedMC carrier blade at RadiSys to scale in smaller steps, and potential reli- (a) is populated with: Corporation. ability problems could be associated with such high call volume on a single slot. n (a) CPU AdvancedMC running a MEGACO or SIP client RadiSys Corporation To meet the smaller densities, lever- n (b) Carrier blade with DSP 5445 N.E. Dawson Creek Drive aging the modularity of AdvancedMC/ AdvancedMC Hillsboro, OR 97124 MicroTCA could be the better choice. In n (c) Carrier blade with optional Tel: 503-615-1100 addition, the media gateway needs to be E1/T1 to the SS7 network E-mail: [email protected] scaled in 8,000 call blocks, which may be n (d) Carrier card with STM-1/OC-3 Website: www.radisys.com cost-prohibitive in many applications. A line cards, connecting to the PSTN mix of AdvancedTCA and AdvancedMC or

MicroTCA and AdvancedMC assets offers (a) (b) a more cost-effective solution. The com- Megaco Processor DSP DSP PQIII AdvancedMC PQIII AdvancedMC bination approaches use AdvancedTCA Carrier Carrier or MicroTCA chassis with AdvancedMC Data Flow cards that handle call volumes of 500 and PPC PPC PPC PPC DSP DSP DSP DSP 10,000 subscribers at one tie. With lower Ctrl Flow, RSYS calls per card, it may not be necessary to RTPRTP or TDM support one-to-one (2N) redundancy. N+1 or N+N may suffice to deliver the required Ethernet Switch Ctrl Flow service quality, thus reducing the overall (c) (d) SS7 portal E1/T1 Line Cards DSP cost of the system and addressing a higher AdvancedMC PQIII STM-1 Carrier PQIII Carrier level of granularity.

4 NP NP NP NP Which solution allows better scaling GbE up and down? FRMFRM FRMFRM Application call volume will determine the approach to take here. Deciding on which of these environments suits your STM-1 Channelized/ application’s market size and your time- Unchannelized to-market deadline must be the first decision made. The market size and call Figure 3

34 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 35 @ www.compactpci-systems.com/rsc TECHNOLOGY

LIQUID COOLING Liquid-cooled embedded computing initiative

By Tahir Cader, Eric Grabowski, Joe Pavlat, and John Peters

In last month’s CompactPCI and is being targeted to achieve a greater totype system was developed to address AdvancedTCA Systems, Tahir, Eric, power density than any competitive air- key technology risks, to prove feasibil- Joe, and John introduced the objectives cooled system, and will possess a number ity, and to provide a reference design. of the Liquid-Cooled Embedded Com- of inherent advantages. Even though the SDM and the prototype puting (LCEC) initiative, a privately systems have some differences, the core funded group tasked with creating an In the early system development stages, engineering development tasks apply to open standard for liquid-cooled the LCEC committee recognized a all LRUs that are completely cooled by embedded computing architectures. number of key technology risks associ- a dielectric coolant, such as Fluorinert. Here they detail development of an ated with implementing liquid cooling. These core tasks include heat acquisition architecture that deploys liquid cooling As a result, it adopted a two-pronged and rejection and hermetic sealing. In for the entire board, also known as full development approach. The first involves addition, these LRUs must have manufac- board cooling. This article continues the development of what is called the turability, reliability, and serviceability/ in full at www.compactpci-systems.com. Static Display Model (SDM). The second maintainability approach leverages a prototype system, The full board cooling solution being developed under a Defense Microelec- Static display model developed consists of a number of tronics Activity (DMEA) program by Figure 1 shows the 9U static display enclosed boards. Each board is envi- member company ISR. The second model subrack with disassembled enclo- sioned to be a Line Replaceable Unit development activity has resulted in a sure and printed circuit board. It is a (LRU) housed in a common chassis. This live system that will simply be referred nonfunctional mechanical model built system, to be discussed in greater detail, to as the prototype system. The live pro- as an example of a potential architecture that could be realized from the standard. Building the SDM has provided critical insight into what it will take to create a practical standard, and is playing an important role as the committee works towards a standard.

Figure 1

The model is intended to roughly represent key subsystems and their placement with respect to each other, and will serve as a baseline as LCEC members gather market- place requirements. The SDM aims to achieve a physical representation of the ideal packaging for a liquid-cooled LRU subrack. c

This article continues in full at www.compactpci-systems.com.

RSC# 36 @ www.compactpci-systems.com/rsc WWW

36 / CompactPCI and AdvancedTCA Systems / October 2006 RSC# 37 @ www.compactpci-systems.com/rsc PRODUCT GUIDE

M i c r o T C A

Clocking in: Real-world MicroTCA needs close clocking/fabric interaction

By Will Chu

In this article, Will briefly surveys MicroTCA architecture Power Modules (PM), and a MicroTCA chassis that comprises and then focuses on four different example systems, with a a backplane, connectors, and cooling units. The AdvancedMCs particular emphasis on the integration of clocks and fabrics deliver specific applications while the remaining elements serve in MicroTCA. The example systems are: as general purpose infrastructure. The MCH serves as the logical and physical hub for delivering IPMI management (the Carrier n General purpose computing and software development in MCH) plus networking of common option and fat pipe fabrics platform and clocking (the Hub in MCH). The PM accepts various AC or n WiMAX basestation DC inputs and provides power conversion, distribution, manage- n 3G/4G basestation ment control, and safety functions. The MicroTCA chassis houses n Quadruple/triple play IP Multimedia System (IMS) platform and cools the AdvancedMCs, MCH, and PM. Figure 1 shows the PICMG MicroTCA RC1.0RC2 block diagram. MicroTCA architecture overview The MicroTCA specification provides system designers with a By intelligently pairing specific types of AdvancedMCs with very flexible architecture. In developing a number of working MCHs that support specific networking protocols and clock MicroTCA systems at CorEdge, we found certain applications types, developers can leverage this generic MicroTCA platform frequently demanded tight coupling of the MicroTCA clocks and to deploy a wide array of applications. Table 1 summarizes a network fabrics, yet the need for this close interaction is often number of possible AdvancedMC/MCH pairings used for differ- misunderstood. ent applications.

MicroTCA systems typically consist of the following elements: For general purpose computing/software development platform AdvancedMCs, MicroTCA Carrier Hubs (MCH), MicroTCA MicroTCA systems, there are a number of implementation

MicroTCA Carrier Power Module #N Cooling Unit #2 MicroTCA Carrier Hub (MCH) #2 Power Module #2 Cooling Unit #1 MicroTCA Carrier Hub (MCH) #1 Power Module #1 MCMC Payload Power Converter Air Mover Management Common Options Fabric Power Converter EMMC Fat Pipe Fabric Clock JSM Power Control EMMC

Backplane Interconnect

AdvancedMC #1 AdvancedMC #2 AdvancedMC #3 AdvancedMC #4 AdvancedMC #5 AdvancedMC #6 AdvancedMC #7 AdvancedMC #8 AdvancedMC #9 AdvancedMC #10 AdvancedMC #11 AdvancedMC #12 Figure 1

Application AdvancedMCs MCH Fabrics MCH Clocks General Purpose Computing/Software Development Processor • Storage • I/O • VGA 1 GbE • SATA/SAS • PCI Express 100 MHz WiMAX WiMAX • Processor • Storage 1 GbE 1 PPS/30.72 MHz 3G/4G DSP • Processor • I/O 1 GbE • Serial RapidIO 8 kHz/19.44 MHz Triple/Quadruple Play (Voice, Video, Data, Wireless) 10 Gigabit Ethernet Processor • 10 Gigabit Ethernet I/O 1 GbE • 10 Gigabit Ethernet 8 kHz/19.44 MHz Table 1

38 / CompactPCI and AdvancedTCA Systems / October 2006 Clocking in: Real-world MicroTCA needs close clocking/fabric interaction

By Will Chu

RSC# 39 @ www.compactpci-systems.com/rsc PRODUCT GUIDE

M i c r o T C A options (Table 2). For our first implementation, we used processor, important to note that AdvancedMCs using PCI Express requires storage, I/O, and VGA AdvancedMCs. A MicroTCA general a single 100 MHz clock source to enable any PCI Express links. purpose computing/software development platform using This is a very tight coupling indeed, and highlights the importance a CorEdge Networks PicoTCA 2UE test and development of having good clocking support in the MCH. platform is shown in Figure 2. In this case, the MCH is supplying a 1 GbE fabric between the processor AdvancedMC, I/O In future implementations (Table 3), an MCH PCI Express fabric AdvancedMC, and the outside world. The MCH supplies a 100 MHz could tie the processor, I/O, and VGA AdvancedMCs into a PCI Express oscillator to serve as the master clock domain between single PCI Express domain with the MCH 100 MHz PCI Express the processor and VGA AdvancedMCs. A direct connection on the oscillator. backplane achieves the PCI Express link between the processor and VGA AdvanceMCs. The MCH supplies a SATA/SAS switch For a MicroTCA WiMAX basestation (Table 4), the key require- fabric between the processor and storage AdvancedMCs. It is ment is support for a 1 PPS clock typically sourced from an external GPS antenna. The MCH generates a 30.72 MHz clock locally that is synchronized to the 1 PPS external clock source and distributes those clocks to the WiMAX AdvancedMCs. Typically the base 1 GbE fabric provided by the MCH has enough bandwidth for the entire system. Some implementations of MicroTCA WiMAX basestations under consideration use a Serial RapidIO fabric in place of the 1 GbE fabric. In this case, an MCH SRIO fabric is required.

For a MicroTCA 3G/4G basestation (Table 5), the key requirement is support for a low latency Serial RapidIO (SRIO) fabric Figure 2 for data transport, 1 GbE fabric for management traffic, and

Modules MicroTCA Carrier Hub/AdvancedMC Backplane Interfaces CLK1 CLK3 Port 0 Port 2 Ports 4-7 Processor AdvancedMC MCH CLK1 1 GbE link to MCH MCH SATA/SAS Fabric Direct connect to VGA Storage AdvancedMC MCH CLK1 MCH SATA/SAS Fabric VGA AdvancedMC MCH CLK1 Direct connect to Processor I/O AdvancedMC 1 GbE link to MCH MCH Local PCI Express 100 MHz clock AdvancedMC 1 GbE Base AdvancedMC SATA/SAS Fabric Base source to AdvancedMC CLK3 Channel A Channel B Table 2

Modules MicroTCA Carrier Hub/AdvancedMC Backplane Interfaces CLK1 CLK3 Port 0 Port 2 Ports 4-7 Processor AdvancedMC MCH CLK1 1 GbE link to MCH MCH SATA/SAS Fabric MCH PCI Express Fabric Storage AdvancedMC MCH CLK1 MCH SATA/SAS Fabric VGA AdvancedMC MCH CLK1 MCH PCI Express Fabric I/O AdvancedMC MCH CLK1 1 GbE link to MCH MCH PCI Express Fabric MCH Local PCI Express 100 MHz clock AdvancedMC 1 GbE Base AdvancedMC SATA/SAS Fabric Base AdvancedMC source to AdvancedMC CLK3 Channel A Channel B PCI Express Fabric Base Channel D, E, F, G Table 3

Modules MicroTCA Carrier Hub/AdvancedMC Backplane Interfaces CLK1 CLK2 CLK3 Port 0 Port 2 WiMAX AdvancedMC MCH CLK2 MCH CLK1 1 GbE link to MCH Processor AdvancedMC 1 GbE link to MCH Direct connect to processor Storage AdvancedMC Direct connect to processor MCH Local 30.72 MHz clock to External 1 PPS GPS clock from face 1 GbE Base Channel A with AdvancedMC CLK3 plate to AdvancedMC CLK2 1 GbE uplinks Table 4

40 / CompactPCI and AdvancedTCA Systems / October 2006 potentially 8 kHz/19.44 MHz telco synchronization clocks Designers also face a number of clocking choices: sourced from an external master system clock. The MCH provides the 1 GbE and SRIO fabric and acts as a clock hub that n 100 MHzDIGITAL clock for PCI MEDIA Express applications SERVERS distributes the incoming telco clocks to the AdvancedMCs. n 1 PPS and 30.72 MHz clock for WiMAX applications n 8 kHz/19.44 MHz clock for telco applications For a MicroTCA quadruple/triple play IMS platform (Table 6), n Custom clocking requirements the key requirement is support for 10 Gigabit Ethernet with advanced flow control and congestion management support To handle this broad range of requirements, CorEdge has taken throughout the system. IMS networks must handle a massive advantage of the multitongue connector architecture in the amount of traffic with precisely controlled quality of service MicroTCA spec to create a series of modular replaceable clock levels. With advanced flow control and congestion management cards for MCH tongue 2 and different fabric MCH cards for dif- support, both the data and management traffic can be carried ferent protocols. This has allowed us to support a wide range across the 10 GbE fabric. This approach obviates the need for of customer application needs, which in turn will be additional fabrics, which reduces overall system costs. In an all- critical for jumpstarting the launching Ethernet system, the requirement for synchronization clocks may of the MicroTCA market. also be relaxed because advanced flow control and congestion Figure 3 shows a base management capabilities coupled with a well-designed packet MCH with PCI Express/ buffering scheme can provide a user experience and network SATA switch module and performance similar to a fully synchronized network. multitongue plug. Figure 3 Implications for MCH design Conclusion In looking at these MicroTCA systems examples, it is clear that The key take-away for companies wanting to develop com- once you have selected your target applications, you select the plete solutions using MicroTCA is to give careful consider- appropriate AdvancedMCs and then select an MCH that supports ation upfront to the interplay between clocks and fabrics in both the application and AdvancedMCs. In a perfect world, a MicroTCA systems. With proper upfront thinking, MicroTCA single MCH would support the entire spectrum of potential appli- applications can be developed relatively efficiently and, in cations that could be addressed by the MicroTCA architecture. many cases, using many off-the-shelf components including However, in the real world that is impossible. Different applica- AdvancedMCs and MCHs. tions require different protocols, different bandwidths, and dif- c ferent clocking schemes. One size MCH cannot fit the complete Will Chu is the president of CorEdge range of requirements. For example, fabric options include: Networks, a developer of AdvancedTCA, MicroTCA, AdvancedMC, and IPMI n 1 GbE as a general purpose fabric for data and products. management traffic n PCI Express fabric for processor AdvancedMCs to other peripheral AdvancedMCs (I/O, VGA, others) To learn more, contact Will at: n SATA/SAS fabric for processor to storage AdvancedMC communications CorEdge Networks, Inc. n Serial RapidIO fabric for Processor and DSP 50 Commonwealth Avenue, Suite 504 AdvancedMCs Boston, MA 02116 n 10 GbE fabric for next-generation processor and I/O Tel: 617-267-5205 AdvancedMCs that will use a 10 GbE backplane interface E-mail: [email protected] n Future fabrics Website: www.coredgenetworks.com

Modules MicroTCA Carrier Hub/AdvancedMC Backplane Interfaces CLK1 CLK2 Port 0 Ports 4-7 DSP AdvancedMC MCH CLK1 MCH CLK2 1 GbE link to MCH MCH SRIO Fabric Processor AdvancedMC MCH CLK1 MCH CLK2 1 GbE link to MCH MCH SRIO Fabric I/O AdvancedMC MCH CLK1 MCH CLK2 1 GbE link to MCH MCH SRIO Fabric MCH External 8 kHz clock to AdvancedMC External 19.44 MHz clock 1 GbE Base Channel A with AdvancedMC SRIO Fabric Base CLK1 CLK2 1 GbE uplinks Channel D, E, F, G w/ SRIO uplinks Table 5

Modules MicroTCA Carrier Hub/AdvancedMC Backplane Interfaces CLK Ports 4-7 10 GbE Processor AdvancedMC with advanced flow control and N/A MCH 10 GbE Fabric congestion management support 10 GbE I/O AdvancedMC with advanced flow control and congestion N/A MCH 10 GbE Fabric management support MCH N/A AdvancedMC 10 GbE Fabric Base Channel D, E, F, G with 10 GbE uplinks

Table 6

CompactPCI and AdvancedTCA Systems / October 2006 / 41 PRODUCT GUIDE

M i c r o T C A

MicroTCA – a new standard for the battlefield

By Rob Persons

Although you may not see specific references to battlefield Developers are adopting Internet Protocol (IP) based strategies applications as you peruse this issue’s MicroTCA Product for WIN-T to maximize throughput and flexibility for new devices Guide, Rob makes the case here for MicroTCA as a viable that are being developed for FCS. Compatibility with the Global architecture for many new military systems such as the Information Grid (GIG) will allow military and civilian leader- Warfighter Information Network – Tactical (WIN-T). ship direct access to the battlefield from around the world. An emerging standard, World Interoperability for Microwave Access The rapid and aggressive transformation mandated by the (WiMAX), which is a certification mark for products that con- Department of Defense for integrated battlefield management, form to the IEEE 802.16-2004 family of standards, along with or the network approach to warfare, has far-reaching influence IEEE 802.11, are two wireless protocols that will be used along over military technology development and insertion. The trans- with other standard IP protocols to improve data throughput in formation of the battlefield to interconnect war fighters to their the field. Sprint has recently announced it will deploy a WiMAX command structure and to other war fighters is driving the need network to support 4G wireless phones for making broadband for new communications strategies and system architectures data performance to the phone possible. that can maximize the amount of Commercial Off-the-Shelf (COTS) content and reuse. Developing high performance mili- Network-centric system architectures tary systems that can leverage large economies of scale has been MicroTCA leverages the emerging ecosystem for AdvancedMCs an elusive reality because most current architectures are tightly to create a new, flexible, small form factor platform, enabling a coupled to the industry that defined the standard. At the same variety of system configurations to be created. The MicroTCA time, maximizing reuse of common nondifferentiating technolo- specification allows for modular or monolithic chassis con- gies, such as CPU cards and disk modules, has been complex figurations from 1 carrier and 1 module to 16 carriers and and difficult. 192 modules, while ensuring that modules always see the same virtual environment. These MicroTCA communications serv- Rapid modernization is only possible when new systems are ers, such as the Motorola Centellis 1000 communications developed around advanced or superior open standard hardware server (Figure 1), typically support 2 to 3 independent fabric and software. Not using COTS solutions has become too expen- interconnects on a carrier, where each fabric port (differential sive and slow to combat the heightening global and homeland transmit/receive pairs) is capable of at least 6.25 Gigabaud security requirements. In addition, resources for modernization (5 Gbps) in each direction, and specific ports can be aggregated are limited, so military and federal agencies are looking for new to form fat pipes with higher throughput. open standard approaches for rapid, cost-effective development and deployment.

The MicroTCA standard based on the Advanced Mezzanine Card (AdvancedMC) was ratified by PICMG in July 2006. Basing MicroTCA on AdvancedMC modules enables the standard’s rapid development and adoption. Companies have a framework to develop new network-centric platforms for small network devices that address both telecommunications and battlefield systems.

Network-centric warfare Figure 1 One effort to modernize the battlefield is a program called Future Combat Systems (FCS), which will improve communi- Systems like the Centellis 1000 will support three different cations and data connectivity between battalion headquarters fabric protocols simultaneously in a chassis: Gigabit Ethernet, and all command structures below, down to the individual sol- PCI Express, and Serial ATA. The current aggregate carrier dier. Radio modernization has been underway for several years (switched backplane) bandwidth is around 40 Gbps, but next through the Joint Tactical Radio System (JTRS) program, where generation hubs should exceed this. The MicroTCA specifica- a single radio will perform a variety of functions by dynamically tion allows for up to 12.5 Gigabaud per port. changing the waveforms. Rear echelon and maneuvering battal- ions of the Future Force (FF) will share voice, data, and video Migration of existing I/O PMCs to the AdvancedMC form factor on the battlefield through a network developed as part of WIN-T is straightforward because PCI Express is an extension of stan- to improve situational awareness for all command structure ele- dard PCI. PCI Express maintains backward software compatibil- ments. The network is being designed to be extremely mobile, ity to PCI so that drivers and operating system software can be resilient, and adaptable for many battlefield situations. reused. PCI to PCI Express bridges can be incorporated on an

42 / CompactPCI and AdvancedTCA Systems / October 2006 AdvancedMC, allowing reuse of existing PMC card designs. As power line filtering, to meet stringent MIL-STD-461 new controllers are released with PCI Express, MicroTCA can EMI/EMC requirements add this new functionality while maintaining more traditional n The ruggedized ATR chassis uses the MicroTCA I/O or custom PCI based designs in the same chassis. Also, specification’s optional locking provisions, to firmly retain migrating to a switched serial architecture like PCI Express the MicroTCA cards into the card cage, providing significant eliminates the slowest link limitations[1]. additional resistance to shock and vibration MicroTCA – a new standard for the battlefield n Military power supply front ends can be used to Applying MicroTCA to WIN-T meet specific military power supply requirements such Referred to as WirelessMAN or Wireless Metro Area Network, as MIL-STD-704 aircraft power or MIL-STD-1275 By Rob Persons IEEE 802.16-2004 was developed to promote wireless broadband vehicle power services to the last mile, the connection from the street to the home. Operating frequencies in the original 802.16 specification The specifications of the AdvancedMC and MicroTCA cards were from 10 GHz to 62 GHz and expanded to include 2 GHz to that are used in a particular application are the limiting factor in 11 GHz frequencies with 802.16-2004, though the frequencies determining the system’s operating temperature range. Commer- will be relegated to licensed frequencies of 2.5 GHz to 2.69 GHz cial AdvancedMC and MicroTCA cards can satisfy less stringent and 3.4 GHz to 3.6 GHz and unlicensed spectrum 5.725 GHz military temperature ranges. Ruggedized and/or extended tem- to 5.850 GHz. In late 2005, 802.16e was added and included perature range AdvancedMC and MicroTCA cards may be Orthogonal Frequency-Division Multiplexing (OFDM), where a required in some applications. This is similar to the existing single transmitter transmits on many (typically dozens to thou- VME and CompactPCI form factors. sands) different orthogonal frequencies, that is, frequencies that are independent with respect to the relative phase relationship Rugged MicroTCA enclosures with adapted AdvancedMC mod- between the frequencies[2]. OFDM improves the performance ules will allow for a more flexible deployment of new network of this wireless protocol, reduces interference between devices, centric technologies to the battlefield. Adoption of PCI Express and allows for Non Line Of Sight (NLOS) operation. The and Gigabit Ethernet as the base fabrics of MicroTCA will reduce IEEE 802.16 standard also utilizes scheduling algorithms that the complexity of migrating critical military I/O to AdvancedMC gives all subscribers a controlled access to the network and form factors and help quickly ramp up the availability of neces- allows for quality of service control of network traffic. sary I/O. New AdvancedMCs that can deliver the new wireless technologies to the battlefield are the same technologies that will Army programs are interested in WiMAX because it has be used in the next generation mobile phones with high-speed been designed to deliver broadband data performance both for broadband capabilities. But it is the efforts of companies investi- fixed and mobile devices. Dynamic creation and maintenance of gating the ruggedization of commercial AdvancedMCs that will these mobile networks along with Quality of Service (QoS) that help the military utilize truly COTS hardware in the battlefield will allow true battlefield situational information between the that is also being deployed in civilian applications. rear echelon and the advancing troops. WiMAX AdvancedMC modules will be available in the future, and MicroTCA will Summary be the platform architecture best suited for deploying it into MicroTCA has the same potential as VME did 25 years ago to the field. become the standard for multiple industries. Many of the telecom- munication industry’s COTS advancements will finally unite with Rugged MicroTCA the network initiatives found in medical, military, and aerospace Working closely with Motorola, Hybricon has developed a to drive true COTS rapid insertion of cost-effective, ubiquitous, ruggedized MicroTCA ATR chassis that leverages Motorola’s high performance, flexible, and scalable platforms. commercial MicroTCA platform, while also accommodating c double-width modules (Figure 2). The ruggedized ATR plat- References form remains compliant with the MicroTCA specification, and [1] Bringing up to PCI Express from PCI, Intel Corporation White Paper, addresses many of the limitations of commercial MicroTCA for http://download.intel.com/design/bridge/papers/25375501.pdf military applications: [2] Orthogonal frequency-division multiplexing, Wikipedia, http:// en.wikipedia.org/wiki/Orthogonal_frequency-division_multiplexing n The ruggedized ATR chassis uses shock isolation of the MicroTCA card cage inside the ruggedized ATR chassis; Rob Persons is a field applications this attenuates the level of shock engineer at Motorola Embedded and vibration that is seen by the Communications Computing. MicroTCA cards, allowing the chassis to meet stringent MIL-STD-810 shock and vibration requirements n Military circular connectors for copper and fiber To learn more, contact Rob at: I/O are used to meet military ruggedization Motorola Embedded Communications Computing requirements for 2900 S. Diablo Way DW205 external connectors Tempe, AZ 85282 n A secondary EMI Tel: 407-699-7129 barrier is used, in E-mail: [email protected] addition to aggressive Figure 2 Website: www.motorola.com/computing

CompactPCI and AdvancedTCA Systems / October 2006 / 43 PRODUCT GUIDE

M i c r o T C A

MicroTCA offers a direct solution for tight cost and size restraint applications

By Stuart Jamieson

Stuart describes the benefits telecom OEMs can realize by taking advantage High-performance roots of the MicroTCA form factor, which In late 2001, under the auspices of PICMG, component suppliers, TEMs, and service providers from enables AdvancedMC modules to be throughout the world came together to define a new open architecture platform for packet-based plugged directly into a backplane, telecom infrastructure applications. The platform adopted in 2003, AdvancedTCA, combined a large without the need for an AdvancedTCA form factor and high power capability with a hot swappable, multiprotocol switched fabric that made it carrier. ideal for packet based telecom systems of high performance, density, and availability. Telecom Equipment Manufacturers (TEMs) and Network Equipment Provid- In 2005, PICMG released a complementary new mezzanine standard for AdvancedTCA. Known as ers (NEPs) are very excited about the new AdvancedMC (PICMG AMC.0), the new mezzanine interface enhanced AdvancedTCA’s flexibility by capabilities provided by AdvancedTCA, extending its high-bandwidth, multiprotocol interface to individual hot swappable modules. Together, AdvancedMC, and MicroTCA, also often AdvancedTCA blades equipped with AdvancedMC modules gave telecom OEMs a versatile platform for referred to collectively as xTCA. TEMs and NEPs must upgrade their basic quickly building modular telecom systems that could be designed, outsourced, manufactured, scaled, architecture to respond to the exploding upgraded, and serviced with a finer degree of granularity at a much lower cost. demand for new telecom, networking, and entertainment services and features, AdvancedMC modules directly into a tocols, including Ethernet, PCI Express/ as well as a tidal wave of new subscribers backplane. No AdvancedTCA carrier AS, and Serial RapidIO. and users. Within a few years, billions of is necessary. This backplane environ- new users from emerging countries like ment, which supports a variety of flex- The foundation for the MicroTCA India and China will also be demanding ible packaging options, enables telecom chassis is the MCH, which provides the these solutions. OEMs to utilize standard off-the-shelf switched fabric, clock distribution, and Advanced modules. What’s more, OEMs shelf management functions needed to Because of this demand, almost every can leverage existing AdvancedTCA/ support up to 12 AdvancedMC modules. major telecom equipment manufacturer AdvancedMC switching, protocol, and See Figure 1. The MCH module acts as and network equipment provider is look- Intelligent Platform Management Inter- a star hub, providing a central switch ing at some flavor of xTCA for the next face (IPMI) infrastructure, thereby help- and high-speed lanes to each module. generation of telecom infrastructure. ing them to preserve their investment in Adding a second MCH to the MicroTCA existing shelf management and applica- enclosure creates the dual star topol- Telecom analyst VDC (Framingham, tion software. This combination of flex- ogy required for reliability. Utilizing Massachusetts) forecasts a $1.15 billion ibility and synergy makes MicroTCA an the same serial transport mechanism as total market for merchant AdvancedTCA excellent complement to AdvancedTCA/ AdvancedMC, the MCH provides a scal- blades by the end of 2009, and a $763 AdvancedMC for small form factor cen- able bandwidth of up to 12.5 Gbps per million dollar market for AdvancedMCs. tral office and outside-plant applications, channel, and an aggregate bandwidth of This bodes well for MicroTCA, which including wireless basestations, digital up to 612 Gbps per MCH. is largely an extension of these two plat- loop carriers, optical ADMs, and Fiber to forms. VDC is already projecting that the the Curb optical network units. MicroTCA backplanes provide redun- market for integrated MicroTCA systems dant I2C-based IPMI, which enable shelf encompassing chassis, power, fans, and MicroTCA architecture management to monitor and control each MicroTCA Carrier Hubs (MCHs) will MicroTCA is in some respects a repack- module installed in the backplane. The grow to about $217 million in 2009. aging of the modular AdvancedTCA/ IPMI Module Management Controller Emerson’s own preliminary market esti- AdvancedMC fabric. In effect, the (MMC) on each AdvancedMC module mates indicate that the total available MicroTCA enclosure acts as a virtual gathers information for temperature, volt- market for MicroTCA platforms will AdvancedTCA carrier, emulating the age, and other parameters that are deemed reach $900 million in 2008, growing to AdvancedTCA carrier environment spec- vital to the module’s normal operation $3.3 billion in 2010. ified in AMC.0. MicroTCA backplanes and conveys them to shelf management. support star, dual star, and full mesh A direct connection topologies. MicroTCA backplane com- MicroTCA shelves will be able to accept MicroTCA reduces cost and size munications are also protocol agnostic, any standard AdvancedMC module in a by enabling telecom OEMs to plug supporting a variety of packet-based pro- variety of form factors, including:

44 / CompactPCI and AdvancedTCA Systems / October 2006 n Half-height/single-wide n Half-height/double-wide n Mid-size/single-wide n Mid-size/double wide n Full-height/single-wide n Full-height/double-wide

A typical high-availability shelf could combine redundant MCHs and power modules with up to 12 AdvancedMC modules. MicroTCA shelves will take power from an AC main or traditional -48 Vdc telecom source, and convert it to 12 V for delivery to individual AdvancedMC modules. Figure 1 Versatile packaging The MicroTCA specification suggests a by 200 mm (roughly 8 inches) deep. tively, MicroTCA enclosures can also be number of packaging options, but does Cubes can be used in a standalone mode. used to build large-capacity systems with not define one as part of the spec. The sug- They can also be assembled into two- hundreds of AdvancedMC modules. gested 19-inch rack-mount MicroTCA dimensional arrays and installed in a chassis, for example, would range from standard rack. Summary 2U to 6U and measure just 300 mm deep MicroTCA is the first open shelf archi- (including cabling), a key requirement MicroTCA enclosures are not limited to tecture to meet the cost, performance, for many optical applications. To accom- standard shelves or cubes. AdvancedMC’s and availability requirements of emerging modate more irregular outside plant, compact size enables MicroTCA enclo- low- to mid-range wireline and wireless pole-mounted environments, work is sures to be used in a variety of space- telecom applications. Equally important, underway to make the MicroTCA chassis constrained applications where only it does so in a way that leverages existing available in a cube configuration, which a few modules (pico assemblies) are AdvancedTCA and AdvancedMC infra- measures 8 inches wide by 8 inches high needed to complete a system. Alterna- structure, enabling TEMS to preserve

RSC# 45 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 45 RSC# 46 @ www.compactpci-systems.com/rsc MicroTCA demonstration PRODUCT GUIDE PICMG has completed putting the finishing touches on a new small form factor open architecture tele- M i c r o T C A com platform that addresses central office, outside their investment in existing software To learn more, contact Stuart at: plant, and last-mile access applications with tight while giving them easy access to off-the- size and cost constraints. Released in July 2006 shelf AdvancedMC modules. Emerson Network Power MicroTCA enables AdvancedMC modules to be used Suite 4, 2 Anderson Place directly in a variety of compact, low-cost enclosures Stuart Jamieson is Edinburgh EH6 5NP UK without the need for an AdvancedTCA carrier. At director of advanced Tel: 44-131-475-7012 technology with Fax: 344-131-475-7001 SUPERCOMM 2005, PICMG performed the first Emerson Network E-mail:[email protected] physical MicroTCA demonstration using a 2U rack Power. Website: www.artesyncp.com based system equipped with five Emerson Pentium M-based AdvancedMC modules.

At the 2006 3GSM show, Emerson took MicroTCA to the next level, demonstrating a turnkey 12-slot MicroTCA system for evaluating and developing wireless base station (WiMAX and 3G), IP Multi- media Subsystems (IMS), MSPP, and IPPBX applications. The system was equipped with KosaiPM payload modules, an MCH module, power supply, fat pipe switch module, application/protocol processing, and platform management software, all developed by Emerson.

At the recent GLOBALCOMM event, Emerson dem- onstrated a variety of multimedia and telecom applications using its new 12-slot MicroTCA development system, which is equipped with a MicroTCA Carrier Hub (MCH) module, power supply, Fat Pipe switch module, application/protocol processing, and platform management software (Figure 2). In one of the demos, Emerson KosaiPM AdvancedMC modules, plugged directly into the MicroTCA chassis, performed application hosting with Surf Communica- tion Solutions AdvancedMC modules handing the IMS video streaming and videoconferencing. In a second demo, a KosaiPM module (Figure 3) running Linux from a hard drive performed videoconferencing.

Figure 2

Figure 3 RSC# 47 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 47 product guide M i c r o T C A

COMPANY NAME DESCRIPTION COMPANY NAME DESCRIPTION Carlo Gavazzi CS www.gavazzi-computing.com MicroTCA Dual Star Single-width, full-height dual star backplane with two Backplane MicroTCA Carrier Hubs (MCHs) and two redundant 6862 Series 1U A complete AdvancedMC and MicroTCA development power modules • 6 AdvancedMCs MicroTCA Starter Kit system • 1.75-inch (1U) (H) x 10.3-inch (W) x 7.7-inch (D) • MicroTCA backplane supports AMC.0, AMC.1, Emerson Network Power www.artesyncp.com AMC.2, AMC.3, AMC.4 • IPMI system manager creates a MicroTCA-like environment KosaiPM AMC card Half-height AdvancedMC for adding control and packet processing performance to small form factor CorEdge Networks, Inc. www.coredgenetworks.com MicroTCA systems and low-profile custom carrier blades equipped with AdvancedMC sites • Hot-swappable, AMC Mechanicals Complete suite of mechanicals for AdvancedTCA Pentium-based card suitable for control plane AdvancedMCs, AdvancedTCA carrier cards, and processor for optical and wireless infrastructure MicroTCA Carrier Hubs • Have passed rigorous NEBS Level 4 shock, vibration, and seismic tests (industry MicroTCA Development Open frame 19" rack-mount chassis with tabletop first) • Available as a suite or a la carte Platform cabinet • 12 single-wide content slots with a combination of 6 full height and 6 half height CEN-MCH MicroTCA Virtual Carrier Manager Controller that has AdvancedMC modules or 9 full height AdvancedMC been proven to support multivendor MicroTCA systems • A MicroTCA Carrier Hub (MCH) module with dual at numerous open interoperability workshops • High front panel 10/100/100 Ethernet ports performance IPMI management and multiprotocol networking support Harting www.harting.com CEN-MCHv2.7 MicroTCA Carrier Hub (MCH) designed to the PICMG AMC pressfit connector AdvancedMC connector for MicroTCA backplanes, MicroTCA v1.0 specification • Serves as the logical and for MicroTCA selected by PICMG for inclusion in the new standard physical hub for delivering management, clock, and • Suitable for routing without the need to use blind fabric connectivity of AdvancedMC-centric MicroTCA or buried vias systems Interphase www.interphase.com CEN-MPWR Power Entry Module designed to the MicroTCA v0.95 specification • Supports an IPMI interface to the iSPAN 3632 AMC MicroTCA ready AdvancedMC for next-generation CorEdge Networks MicroTCA Carrier Hub media gateways, basestation controllers, and radio network controllers • Quad-Port channelized CEN-MTCv1.0 Small, self-contained chassis-based test system that single-width, single-height module OC-3/STM-1 supports testing of multiple AdvancedMCs • Supports interface processor used to support integration with 2 AdvancedMCs with integrated shelf management TDM telecom interfaces in the network • Supports 1 MicroTCA Carrier Hub and 1 AdvancedMC to simulate a MicroTCA system iSPAN 36NP AMC MicroTCA ready network processor card is a single- width, single-height AdvancedMC packet processing CEN-PICO-1UR PicoTCA/MicroTCA/AdvancedMC development engine designed for use with deep packet inspection, system with RTM customizable to support various packet forwarding, and packet switching applications I/O interfaces and provide I/O access to all • Intel IXP2350 packet processor 21 AdvancedMC ports • MicroTCA midplane supports AMC.0, AMC.1, AMC.2, AMC.3, AMC.4 specifications Kaparel www.kaparel.com for one AdvancedMC (single or double wide) AdvancedTCA Enclosure Systems up to Level 5 for AdvancedTCA and MicroTCA CEN-PICO-iUS PicoTCA/MicroTCA/AdvancedMC 1U standalone test Family • Everything is fully assembled, ready to run and and development system • Cross-connected MicroTCA individually configured • Case solutions in 5U, 12U, backplane supports one double wide or two single 13U, or cube design wide AdvancedMCs and AMC.0, AMC.1, AMC.2, AMC.3, Linear Technology www.linear.com AMC.4 specifications LT4351 MicroTCA MOSFET diode OR controller • Low-cost MicroTCA MCH Hub MicroTCA Carrier Hub (MCH) card that controls multiple replacement for ORing diode in multiple power supply AdvancedMC cards in a MicroTCA chassis • Flexible applications front panel I/O options including telco alarms, Ethernet, and various serial interfaces LTC4221 MicroTCA dual hot-swap controller with a dual-level circuit breaker • Allows safe board insertion and MicroTCA MCH Hub v3 MicroTCA Carrier Hub (MCH) card that controls removal from a live backplane multiple AdvancedMC cards in a MicroTCA chassis • High performance IPMI management and Motorola www.motorola.com/computing multiprotocol networking support Centellis 1000 MicroTCA open application-enabling platform ELMA Bustronic www.elmabustronic.com • Integrated and verified hardware and software components • Physically smaller, with finer-grained 14-Slot MicroTCA 14-slot backplane that complies to MicroTCA.0 Rev 1.0 scalability than communications servers based on • High speed connector up to 6.25 Gbps and slot-to-slot the AdvancedTCA industry standard aggregate bandwidth of 5,000 MBps Performance Technologies www.pt.com ELMA Electronic www.elma.com AMC111 Single width, full-height single board compute module MicroTCA 4U System 19" rack-mount, compliant to MicroTCA 1.0 • Single for AdvancedTCA and MicroTCA systems • 64-bit single star backplane accepts 12 AdvancedMC modules core AMD Turion 2.0 GHz processor MicroTCA 7U Cube 19" rack-mount, compliant to MicroTCA 1.0 Draft 0.9 AMC131 Single width, full-height 32-bit AdvancedMC compute Development System • Dual star backplane • 2 MicroTCA Carrier Hubs (MCH) module designed for AdvancedTCA and MicroTCA • 2 redundant power modules • 6 AdvancedMCs, Single systems • Freescale Dual Core 1 GHz vMPC8641D width/full height PowerPC Processor MicroTCA 7U System 19" rack-mount, compliant to MicroTCA.0 Rev. 1.0 • Single star backplane accepts 12 AdvancedMC modules

48 / CompactPCI and AdvancedTCA Systems / October 2006 product guide M i c r o T C A

COMPANY NAME DESCRIPTION PMC-Sierra www.pmc-sierra.com PM6352 RSE 160 Carrier-class 16-port Serial RapidIO switch capable of scaling up to 10 Gbps per port that is suitable for AdvancedTCA- and MicroTCA-based wireless infrastructure platforms • Protection and diagnostic features for carrier grade applications Positronic Industries www.connectpositronic.com MicroTCA Input The “QB” Series meets requirements of the MicroTCA Power Connectors Specification for 48 V and 24 V systems • Board mount connectors for power modules and cable connectors for bringing power to modules. Rittal www.rittal.com MicroTCA system MicroTCA system that accommodates AdvancedMC cards • 2U, 3U, and 5U (high) configurations • Built-in fan aids shelf cooling to keep AdvancedMCs cool in 3U and 5U versions SANBlaze Technology www.sanblaze.com SB-AMC-HD A drive carrier module for AdvancedTCA or MicroTCA systems • Configured with either one 2.5 inch SATA or one SFF SAS drive Schroff www.schroff.us 6U MicroTCA 6U, 316 mm deep ratiopac PRO case with front handles Development System • Card cage for single width, full-height AdvancedMC modules • 14-slot MicroTCA backplane (two power module slots, two MCH slots, 10 AdvancedMC slots) or 16-slot (two power module slots, two MCH slots, 12 AdvancedMC slots) 8U MicroTCA 8U, 316 mm deep ratiopac PRO case with front handles Development System • Card cage for double width, full-height AdvancedMC modules • 14-slot MicroTCA backplane (two power module slots, two MCH slots, 10 AdvancedMC slots) or 16-slot (two power module slots, two MCH slots, 12 AdvancedMC slots) Tyco Electronics Power Systems www.tycoelectronics.com MicroTCA Backplane Press-fit through-hole and surface mount configurations Connector • Designed for high-speed differential pair applications (12.5+ Gbps) MicroTCA Power 24 individual 15 Amp power contacts • 72 individual Connector high density signal contacts • All contacts stamped and formed Yamaichi Electronics www.yeu.com AdvancedMC BackPlane Compression technology reduces number of layers Connector • Highest signal speed capability beyond 12.5 Gbps with GR-1217-CORE compliant reliability CN080 Series Connector Compliant to MicroTCA with highest speed perfomance beyond 12 Gbps • GR-1217-CORE and RoHS compliant VCM Plug Connector MicroTCA and GR-1217-CORE compliant • Available in 1, 2, 3, or 4 (all) tongues • Solves tolerance stack up problem between aggregate backplane connectors

RSC# 49 @ www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / October 2006 / 49 RSC# 50 @ www.compactpci-systems.com/rsc RSC# 51 @ www.compactpci-systems.com/rsc RSC# 52 @ www.compactpci-systems.com/rsc ® OpenSystems CompactPCI Publishing™ OpenSystems Publishing ® Advertising/Business Office and 30233 Jefferson Avenue St. Clair Shores, MI 48082 AdvancedTCA Systems Tel: 586-415-6500 n Fax: 586-415-4882 Vice President Marketing & Sales ADVERTISER INFORMATION Patrick Hopper [email protected] Page/RSC# Advertiser/Product description Business Manager 9 ADLINK Technology – AdvancedTCA Products Karen Layman ® 47 Advanet – CompactPCI, VME, PMC Series CompactPCI Sales Group 52 Arrow Electronics – MicroTCA Communication Servers ® Dennis Doyle 16 Az-Com – Quality Extenders Senior Account Manager and [email protected] 49 A Concurrentdvan –ce Intel BaseddTCA SBCs Systems Tom Varcie Account Manager 2 Conec – AdvancedTCA Zone 1 Power Connectors [email protected] 7 Diversified Technology – ATCA Blades Doug Cordier Account Manager 37 Elma Bustronic – PICMG Compliant Backplane [email protected] Barbara Quinlan 15 Elma Electronics – Handles and Panels Account Manager [email protected] 39 Embedded Planet – System Creation Andrea Stabile 3 Emerson Network Power – Communication Products Advertising/Marketing Coordinator [email protected] 13 Excalibur Systems – Avionics Communications Christine Long E-marketing Manager 46 GE Fanuc Embedded Systems – SBCs [email protected] 5 GE Fanuc Embedded Systems – Embedded Products 33 Hartmann Elektronik – CompactPCI Backplanes Regional Sales 21 Kontron – Open Modular Solutions Jane Hayward Regional Manager – California 20 N.A.T. – µTCA, MCH, AMC products [email protected] Phil Arndt 23 National Instruments – High-Speed Digital I/O Regional Manager – East Coast [email protected] 24 One Stop Systems – CompactPCI Express Richard Ayer 11 Performance Technologies – CPC5564 64-Bit AMD Opteron SBC Regional Manager – West Coast [email protected] 50 Performance Technologies – Advanced Managed Platforms 35 Positronic Industries – MicroTCA Input Power Connectors International Sales 25 Radian Heatsinks – Low profile heat sinks Stefan Baginski European Bureau Chief 51 RadiSys Corporation – AdvancedTCA Solutions [email protected] Dan Aronovic 36 Red Rock Technologies – Mass Storage Modules Account Manager – Israel [email protected] 27 Schroff – Hybrid Serial Parallel Cooling 45 Simon Industries – Conduction-Cooled Heat Frames Reprints and PDFs 6 TEWS Technologies – Embedded I/O Solutions Call the sales office: 586-415-6500 17 Winchester Electronics – Signal and Power Connectors

www.compactpci-systems.com/rsc RSC# 54 @ www.compactpci-systems.com/rsc