Compactpci and Advancedtca Systems

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

RSC# 2 @www.compactpci-systems.com/rsc RSC# 3 @www.compactpci-systems.com/rsc table of contents

CompactPCI and PERF TECH CompactPCI AdvancedTCA ® Systems is and AdvancedTCA Systems published by: OpenSystems Volume 8 / Number 8 / November 2004 PublishingTM

COLUMNS FEATURES

Editor’s Foreword 8 Un-tuned 22 SPECIAL FEATURE: Embedded Software By Joe Pavlat Manageability in AdvancedTCA and the role of IPMI By Sanjoy Maity, American Megatrends Inc. AdvancedMC Series 10 Advanced Mezzanine Card The role of standards based high availability (AdvancedMC) design goals middleware in accelerating product development By Lawson Guthrie and Mark Summers By John Fryer, Motorola Embedded Communications Computing Group PCI Express 14 PCI Express: Legacy PCI software vs. PCI Express-aware software Best practices for secure console port management: By Chris Youman and Steve Cooper Establish a virtual crash cart to minimize downtime every time Software Corner 18 By Charles Pheterson, Equinox Sun Microsystems announces AdvancedTCA plans at ITU Telecom Asia By Curt Schwaderer 35 APPLICATION FEATURE: Hot Swap CompactPCI hot swapping StarFabric Watch 20 By Alexey Polonsky, Jungo Software Technologies StarFabric picks up the pace By Pat Hart 38 TECHNOLOGY FEATURE: Military New Products 49 Israel Defense Forces deploy AdvancedTCA core routing switch By Zvi Marom, Telco Systems

40 TECHNOLOGY FEATURE: ON THE COVER: High performance Can an AdvancedTCA based platform Building high performance handle core routers/switches required to support line cards with 10 Gbps edge/core switches using an user traffic? TeraChip is building high AdvancedTCA platform performance edge/core switches using an AdvancedTCA platform – implementation issues and discusses implementation By Reuven Segev, in this issue. TeraChip, Inc.

4 / CompactPCI and AdvancedTCA Systems / November 2004 PERF TECH

RSC# 5 @www.compactpci-systems.com/rsc product guide CompactPCI ® Analog • Digital • Industrial • Multifunction ® and AdvancedTCA Systems The Magazine for Developers of Open e v e n t s Communication, Industrial, and Rugged Systems An OpenSystems Publication Internet Telephony Miami, Florida ADVERTISING/BUSINESS OFFICE February 22-25 30233 Jefferson Avenue St. Clair Shores, MI 48082 e-letters online Tel: 586-415-6500 • Fax: 586-415-4882 EDITORIAL/PRODUCTION OFFICE NOVEMBER E-LETTER: 13253 La Montana, Suite 207 Fountain Hills, AZ 85268 On the right track: A telecom grade Tel: 480-967-5581 • Fax: 480-837-6466 Application Enabling Platform forms a key element of a rail trafﬁc management system PUBLISHERS John Black, Michael Hopper, Wayne Kristoff By Mark Crampton and Shreekant Raivadera, Motorola Embedded Communications Computing Group EDITORIAL DIRECTOR Joe Pavlat [email protected] PRINT ISSN #1098-7622 ONLINE ISSN #1550-0381 ASSOCIATE EDITOR CompactPCI and AdvancedTCA Systems is published monthly, except bi-monthly Anne Fisher [email protected] in January/February, May/June, and July/August by OpenSystems Publishing LLC., 30233 Jefferson Ave., St. Clair Shores, MI 48082. SENIOR EDITOR Subscriptions are free, upon request in writing, to persons dealing with or consid- Terri Thorson [email protected] ering CompactPCI technology. For others inside the US and Canada, subscriptions are $24/year. For 1st class delivery outside the US and Canada, subscriptions are TECHNOLOGY EDITOR $90/year (advance payment in US funds required). Periodicals postage is paid at Curt Schwaderer Columbus, WI and an additional mailing office. [email protected] POSTMASTER: Send address changes to NEW PRODUCTS EDITOR CompactPCI and AdvancedTCA Systems Chad Lumsden OpenSystems [email protected] 13253 La Montana, Suite 207 TM Fountain Hills, AZ 85268 Publishing SENIOR TECHNICAL EDITOR Mark K. Barrera

NEWS sales office [email protected]

VP OF MARKETING & SALES ACCOUNT MANAGER CONTRIBUTING WRITERS Patrick Hopper Tom Varcie Steve Cooper, Lawson Guthrie, Pat Hart, [email protected] [email protected] Mark Summers, Chris Youman SENIOR ACCOUNT MANAGER EUROPEAN REPRESENTATIVE Dennis Doyle Stefan Baginski +41 (0)62 758-3222 MANAGING EDITOR Bonnie Crutcher [email protected] [email protected] ACCOUNT MANAGER MARKETING REPRESENTATIVE VICE PRESIDENT, EDITORIAL Doug Cordier Andrea Stabile Rosemary Kristoff [email protected] [email protected] [email protected]

PRINT/ONLINE MARKETING SPECIALIST FOR REPRINTS: ART DIRECTOR Christine Long Call the Sales Office Stephanie Sweet [email protected] [email protected] SENIOR WEB DEVELOPER Konrad Witte

SUBSCRIPTIONS: For new subscribers or to change an address WEB DEVELOPER Eric Okorie go to www.opensystems-publishing.com/subscriptions CIRCULATION/OFFICE MANAGER Phyllis Thompson ® CompactPCI, PICMG, AdvancedTCA, ATCA, and their logos are registered trademarks of the [email protected] PCI Industrial Computer Manufacturers Group. BUSINESS MANAGER ® CompactTCA is a trademark of the PCI Industrial Computer Manufacturer’s Group. Karen Layman © 2004 CompactPCI Systems

6 / CompactPCI and AdvancedTCA Systems / November 2004 RSC# 7 @www.compactpci-systems.com/rsc EDITOR’SEDITOR’S FOREWORDFOREWORD By Joe Pavlat Editorial Director Un-tuned CompactPCI & AdvancedTCA

Consumers have demonstrated that they like wireless technology. spectrum, which is generally unavailable, or higher power, which There is something liberating about making a phone call from is undesirable in battery powered hand-held devices. just about anywhere or accessing the Internet over a Wi-Fi connection. But many consumer electronic devices, including DVD UWB uses a much different approach. It works by simultaneously players, high definition televisions, video camcorders, and MP3 transmitting data at very low power levels over a very wide range players are still interconnected with a jumble of cables and wires. of frequencies. This is called underlay access. Because it trans- This may change in the near future, however, as a relatively new mits at a very low power level at any given frequency, conven- technology called Ultra Wide Band (UWB) begins to take shape. tional radio receivers see UWB signals as nothing more than a low Both manufacturers of consumer electronics as well as semi- level of background noise. In 2002 the Federal Communications conductor manufacturers are moving forward to commercialize Commission became convinced and ruled that UWB can operate the technology at a fast pace, but there are still some potholes in without license in a range of 3.1 GHz to 10.6 GHz. Of course, the road ahead. other types of communications including satellite broadcasts, radar, and microwave transmissions use these frequencies. UWB Familiar wireless technologies like cellular telephony, Bluetooth, devices already on the market deliver useful data rates of over and Wi-Fi (802.11b) work by modulating a single carrier fre- 100 Mbps at distances of up to 10 meters and promise future rates quency. Cellular networks use very expensive, licensed frequen- of over 1 Gbps over ranges of a meter or two. cies, and Wi-Fi and Bluetooth use chunks of unlicensed – sometimes called garbage – spectrum. Each technology works well With the enormous consumer electronics industry driving the but lacks sufficient data transfer rate to tackle video oriented con- demand for UWB technology, UWB transceivers will likely sumer devices. Increasing data rates requires either more radio become very cheap very quickly. This may have interesting

RSC# 801 @www.compactpci-systems.com/rsc RSC# 802 @www.compactpci-systems.com/rsc

8 / CompactPCI and AdvancedTCA Systems / November 2004 EDITOR’SEDITOR’S FOREWORDFOREWORD implications for the embedded computer industry. Already the data to be transmitted with pseudo-random numbers to create a need for gigabit data rates is causing the computer industry to broadband signal that looks like noise over the entire frequency change from parallel data buses to switched serial interconnects, range. Working silicon implementing DS-UWB is already avail- sometimes called fabrics. At these high data rates, boards and able. The competing technology, called Multi-Band Orthogonal backplanes must be designed very carefully. Often expensive Frequency Division Multiplexing (MB-ODFM) works slightly and single sourced high-speed connectors are required for com- differently and isn’t compatible with the other approach. munications between boards in a system. Anyone who has been MB-ODFM divides the entire frequency range into 15 bands and involved with PICMG or VITA over the last few years also knows frequency hops between them. The signal processing required of the intellectual property issues and licensing hassles that often to decode signals is somewhat more complex with this tech- surround these connectors. nology, but silicon should be available shortly. Each group has attempted to cement their standard through the IEEE, which has If UWB chips cost only a few dollars and can deliver data rates approved each technology at different times, but each currently in excess of 1 Gbps over short distances, this technology might lacks the requisite 75 percent approval for IEEE endorsement. replace wired interconnects in some future embedded systems. It is likely that each group will now attempt to create either an In addition to the possibility of board-to-board data transfers, it industry standard or de facto standard established by competi- is easy to imagine box-to-box communications being enabled by tion in the marketplace. Either approach will delay the broad simple physical proximity. This has interesting implications for commercial rollout of UWB, but the technology is too compel- scalability and maintainability. Hot swapping a defective board ling to ignore. could be as simple as just managing the power connection. Fewer electrical interconnections should increase reliability as well. So, stay tuned. Or, in this case, un-tuned.

There is a problem that may delay adoption, however. Not unlike the old VHS versus Betamax battle, there are two strong industry initiatives that take slightly different approaches to providing UWB communications. The UWB Forum supports something called Direct Sequence UWB (DS-UWB), which combines the Joe Pavlat, Editorial Director

RSC# 9 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 9 ADVANCEDMCADVANCEDMC SERIESSERIES By Lawson Guthrie Advanced Mezzanine Card and Mark Summers CompactPCI & (AdvancedMC) design goals AdvancedTCA

This is the latest in a series of articles Limitations of existing mezzanines options will be explained further in the describing Advanced Mezzanine Card, The emergence of the AdvancedTCA next article in this series. the industry’s next generation mezzanine specification and high-speed serial inter- standard, by providing a preview of the connect technologies have exposed some Scope of the specification PICMG AdvancedMC.0 specification. shortcomings of existing mezzanine stan- The PICMG AdvancedMC.0 specifica- In this article Lawson and Mark dards, which are subject to bus through- tion defines the baseline requirements for describe why we need a new put limitations and do not enable hot a wide spectrum of high-speed mezzanine mezzanine standard. swap. The advent of highly dense, modu- cards. It includes the common mechanical lar, scalable, and highly reliable network aspects as well as hot swap, power, ther- A mezzanine card, referred to as a mod- infrastructure equipment requires a new mal, and interconnect system management ule within the AdvancedMC specifica- mezzanine standard designed from the elements required for each implementation, defines an add-on or child card that outset to support hot swap capability, tion, In addition PICMG AdvancedMC.0 is intended to extend the functionality of a Intelligent Platform Management Inter- covers regulatory guidelines. carrier board, providing system designers face (IPMI) based manageability, and with the flexibility to enhance their solu- scalable throughput for the next 10 to 15 The specification provides the definitions tions by adding modular building blocks. years. It should be noted that in addition to and requirements for face plates with The benefits include the flexible imple- supporting AdvancedTCA, AdvancedMC ejectors, defined component spaces, com- mentation of unique functions to reach can also support other form factors, plete mechanical dimensions, thermal new and growing market segments, with including proprietary designs. definitions, mounting, guides, and a con- important economies of scale that can nector necessary to interface between the help to reduce equipment costs. Designed to meet telecom Reliability, module and the carrier board. Availability, and Serviceability (RAS) As we have noted in previous articles in requirements, the AdvancedMC feature set The PICMG AdvancedMC.0 specifica- CompactPCI and AdvancedTCA Systems, as defined in the PICMG AdvancedMC.0 tion also defines the framework or base mezzanine cards have evolved steadily specification includes support for single- requirements for a family of anticipated with the ongoing development of modular width and double-width modules in subsidiary specifications that define the bus standards. In the early 1990s, a base addition to full-height and half-height usage requirements for various inter- mezzanine standard was developed to modules. Figure 1 shows four single- faces, including PCI Express, Advanced support parallel interfaces and was opti- width AdvancedMC modules on an Switching, Serial RapidIO, and Gigabit mized to support PCI and CompactPCI AdvancedTCA carrier board. AdvancedMC Ethernet. These and other interconnect environments. The first of these mez- also provides flexible supports for single definitions will be provided through zanine specifications was IEEE 1386, layer and stacked modules. AdvancedMC.x subsidiary specifications. known as the Common Mezzanine Card (CMC) specification. Many subsidiary In addition to these multiple module con- As shown in Figure 1, each AdvancedMC specifications to CMC have subsequently figurations, some of the other notable module lies parallel to the carrier board, emerged, including PCI mezzanine card, improvements enabled by PICMG PCI telecom mezzanine card, and VITA 32 AdvancedMC.0 include: processor PMC. Telecom and server equipment manufacturers have widely adopted ■ Support for hot swap capability these standards. ■ Greater throughput (21 duplexed LVDS ports, each capable of Mezzanines cover a broad range of func- 12.5 Gbps speeds) tionality, but they typically include the ■ An enhanced power budget following general usage models: (up to 60 W per connector) ■ Greater size (25 mm deeper than ■ Telecom connectivity (ATM/POS today’s CMC standard) [OC-3/12/48], T1/E1, VoIP, Gigabit ■ IPMI based system management Ethernet) ■ Processors (CPUs, Digital Signal These enhancements make AdvancedMC Processors, and FPGAs) a compelling mezzanine alternative for ■ Network Processor Units (NPUs) increasing the density and flexibility of ■ Network communications AdvancedTCA and proprietary carrier coprocessors (classification, security, board designs used in today’s demand- or intrusion detection) ing telecom and high-density bladed ■ Mass storage server applications. Various configuration Figure 1

10 / CompactPCI and AdvancedTCA Systems / November 2004 CompactPCI and AdvancedTCA Systems / November 2004 / 11 RSC# 11 @www.compactpci-systems.com/rsc ADVANCEDMCADVANCEDMC SERIESSERIES plugging into the carrier by means of an PICMG 3.0 base specification and count, thereby reducing the amount AdvancedMC connector. Carrier boards build upon its strengths of RAS. The of space required on both the module themselves range from passive boards AdvancedMC module should not be and the carrier board, yet provide with minimal intelligence to high-perfor- limited by other chassis standards. sufficient real estate for intended mance SBCs. ■ System management: System man- interconnects and usage models. agement should be an extension of ■ Support for a rich mix of processors: Why we need a new mezzanine the PICMG 3.0 shelf management This includes CPUs, NPUs, DSPs, Rather than providing another extension Intelligent Platform Management Bus and I/O processors. of the CMC standard, a new mezzanine (IPMB) scheme. ■ Reduced development time and costs: standard was required to meet the design ■ Hot swap support: Hot swap of The reduced total cost of ownership objectives of high-speed serial LVDS AdvancedMC modules should be and shorter design cycles are sup- interface support and AdvancedTCA enabled to support availability and ported by component standardization optimization. As such, AdvancedMC is serviceability objectives. The focus is and economies of scale. not backward compatible with mezzanine on front-loadable hot swap modules. ■ Communications and embedded standards based on the CMC specifica- The PICMG AdvancedMC.0 revi- industry focus: Target usage models tion. AdvancedMC is also adapted to the sion is hot swappable only. While include support for edge, core, increased power budget and size enabled non-hot swap will be an optional transport, data center, wireless, wire- by AdvancedTCA. implementation, it is not addressed in line, and optical network design this revision. Non- hot swap, which elements. Design goals will be covered in the next revision of ■ Modularity, flexibility, and con- This PICMG AdvancedMC.0 specifi- the specification, enables lower cost figurability: PICMG AdvancedMC.0 cation was written to help the telecom implementations. provides support for a minimum industry meet a series of important design ■ LVDS interconnect: AdvancedMC of four modules across a given goals: should be optimized for LVDS inter- AdvancedTCA carrier, and includes connects. dual-width and stacked mezzanines. ■ PICMG 3.0 optimized: All elements ■ Low pin count: The interconnect ■ Future advances in signal throughput: must work within the bounds of the should be conservative in its total pin Anticipates advances in interconnect

RSC# 12 @www.compactpci-systems.com/rsc

12 / CompactPCI and AdvancedTCA Systems / November 2004 technologies by supporting a mini- Mark Summers is For further information, contact either mum of 12.5 Gbps throughput per a technical market- Lawson or Mark at: LVDS signal pair. ing engineer in the Intel Embedded Lawson Guthrie Summary Architecture Divi- Intel Corporation AdvancedMC represents the industry’s sion developing 15400 Greenbrier Pkwy. next generation mezzanine standard, new markets for Beaverton, OR 97006 designed to enable designers to take Intel products and Tel: 503-264-1697 advantage of the performance, scalabil- technologies. As chairman of the PICMG E-mail: [email protected] ity, and cost benefits of AdvancedTCA. AdvancedMC.0 subcommittee, Mark is Website: www.intel.com In addition to providing hot swap capa- focused on assuring a successful industry bility, intelligent platform management, specification that will be readily adopted Mark Summers greater potential power/size, and the by industry. During his 19 years of techni- Intel Corporation other enhancements made possible by cal experience (as employee of Motorola 5000 W. Chandler Blvd. AdvancedTCA, AdvancedMC is intended and Intel) spanning commercial, indus- Chandler, AZ 85226 to help designers take advantage of the trial, and military electronics markets he Tel: 480-554-1258 performance enhancements of new, fast, has been issued 20 US Patents and has E-mail: [email protected] wide-bandwidth serial buses, including authored numerous technical journals. Website: www.intel.com PCI Express.

D0.97 of the PICMG AdvancedMC.0 specification is in the process of Negative Ballot Review and IP Review. The Technical Membership Review process provides PICMG members with the opportunity to submit issues with the specification. IP Review lets companies disclose any intellectual property con- flicts and satisfy outstanding licensing requirements.

Telecom equipment manufacturers should take a close look at the value of AdvancedMC in their next generation designs. Visit www.picmg.org for further information and to download the PICMG AdvancedMC.0 short form specification.

In the next article in this series, we will continue our preview of the PICMG AdvancedMC.0 specification by examining the theory of operations of AdvancedMC modules.

Lawson Guthrie is a strategic initiatives manager in Intel’s Communications Infrastructure Group. He currently serves as secretary of the PICMG AdvancedMC.0 base specification and AdvancedMC.2 (GbE) subsidiary specification, as well as chair of the AdvancedMC.1 (PCI Express and Advanced Switching) subsidiary specification. Lawson has held various positions in strategic marketing over his 17-year career, focused on technology definition for network operating systems, desktop management, and telecommunications. RSC# 13 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 13 PCIPCI EXPRESSEXPRESS By Chris Youman PCI Express: Legacy PCI software and Steve Cooper CompactPCI & vs. PCI Express-aware software AdvancedTCA

Perhaps one of the most compelling ber, device number, and function num- Error handling aspects of PCI Express technology is its ber to PCI Express devices. PCI Express Baseline error handling is the minimum adherence to the legacy PCI software features supported by this compatible amount of error reporting required by and configuration model. Switches and configuration model include baseline all PCI Express devices to be compat- bridges that support an advanced serial error reporting, power management, hot ible with PCI. Software within the PCI- protocol and operate seamlessly in sys- plug, and other features common to both compatible configuration region handles tems running conventional PCI software PCI and PCI Express. PCI Express-defined configuration of a PCI Express device for create a large reason for the momen- features, such as advanced error reporting, supporting baseline error mechanisms. tum driving PCI Express. This column quality of service (traffic classes and virtual will illuminate which functions in PCI channels), and others, require PCI Express- The baseline error handling mechanism Express can be used in legacy PCI OS aware software and enhanced PCI Express determines error type (correctable or environments and which features require configuration space in the device (Table 1). uncorrectable) but does not identify the PCI Express-aware software. Enhanced PCI Express configuration space severity of the error as in the PCI Express exists in the range defined by PCI Express advanced error reporting structure. PCI Express as a standard CPU-to-I/O above 256 bytes up to 4 KB. PCI Express errors can occur at any communications bus performs the same function as the PCI bus but at higher speeds. This architecture assumes the system contains only one main CPU. This CPU controls everything from basic power up to initialization to running the operating system. It is possible to have other CPU components providing intelligent I/O or secondary data processing, but one main CPU always provides system-level enumeration, configuration, and control.

By specification, all PCI Express devices such as root complexes, switches, bridges, and endpoints contain a configuration mechanism that is a superset of that of PCI; the first 256 bytes of device configuration space are the same in PCI Express as in any legacy PCI device. This is possible Figure 1 because the hierarchical bus, device, and function mapping are defined the same way PCI Express as replacement PCI Express with new in both PCI and PCI Express (see Figure 1). for PCI feature support For example, PCI Express switches are configured with an internal PCI bus num- BIOS Will need rewrite since motherboard Will need rewrite to take ber between ports that is distinct from the chipset is different, but PCI-specific advantage of new features. primary PCI bus mapped to its upstream code is unchanged. port and subordinate PCI buses mapped OS No change. Will need rewrite to take to each downstream port. Therefore, the advantage of new features. system software treats switches and mul- tiport PCI Express devices as if they were Drivers Will need rewrite since I/O board Will need rewrite to take a hierarchy of PCI-to-PCI bridges. As the chips are different, but PCI-specific advantage of new features. host enumerates the system, it walks the code is unchanged. bus in much the same way as in PCI and Networking No change. No change. assigns bus, device, and function numbers to every PCI Express port. Applications No change. Will need rewrite to take advantage of new features. Legacy PCI software will configure and assign a unique combination of bus num- Table 1

14 / CompactPCI and AdvancedTCA Systems / November 2004 RSC# 15 @www.compactpci-systems.com/rscCompactPCI and AdvancedTCA Systems / November 2004 / 15 PCIPCI EXPRESSEXPRESS level in the protocol: physical layer, data error types, depending on system prefer- as Advanced State Power Management link layer, or transaction layer. Messages ences. (ASPM). ASPM deals with various power can be signaled as part of a completion saving modes in a PCI Express hardware transaction or through in-band messaging. Power management device through hardware designed state Legacy software maps errors to PCI error The support for PCI Express power man- machines. While this function is trans- mechanisms, including parity and system agement registers exists in the PCI-com- parent to software, it requires software errors. Advanced error mechanisms in patible region of device configuration that is aware of this feature, for example, PCI Express also determine the severity space. By specification, PCI Express PCI Express-aware in order to be properly of errors according to descriptions in power management is compatible with the utilized at the system level. the enhanced PCI Express configura- PCI specification, and no new software is tion space and thus require PCI Express- required to support this in PCI Express Hot plug aware software to function. Such soft- components. However, PCI Express does Similarly, PCI Express hot plug regis- ware also can choose to mask certain provide the additional feature known ters exist in the PCI compatible region of device configuration space. By specification, PCI Express uses the same model as the PCI standard hot plug specification. However, to accurately utilize PCI Express hot plug features, software must use extended PCI Express configuration space within PCI Express devices. This can be done with updated drivers that function within legacy operating systems. The interrupts and other service mechanisms that are associated with hot plug support will be handled according to the operating system.

Quality of service Differentiated classes of traffic and virtual channel support are optional features in PCI Express. These features require PCI Express-aware software to be configured and utilized. Configuration is supported in the extended space within a PCI Express device but not supported with legacy PCI software, as the PCI protocol does not define such features. The configuration space requires specific descriptions in order to describe the capabilities of each port’s virtual channel and traffic class structures, as well as how each of these features will be managed. For example, up to three different arbitration schemes for prioritizing traffic among several virtual channels can be utilized; it is up to the software to understand and manage the status and control of such schemes for each PCI Express device and port in the system.

With an understanding of PCI and PCI Express software compatibility, board and system makers can realize an efficient and practical path to elevating their next-generation designs interconnect plateau.

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

16 / CompactPCI and AdvancedTCA Systems / November 2004 Chris Youman is a senior marketing manager for PLX INDUSTRY NEWS Technology, Inc. Prior to PLX, Chris held sales positions at LSI Logic, where he was a major For the latest industry news, account salesman on telecommunications and data networking accounts. visit our newswire at Prior to that, he was at Advanced Micro Devices, where he was a marketing www.compactpci-systems.com and manager for embedded processors and a global account sales manager on the www.advancedtca-systems.com. IBM account. He holds a BSEE degree from University of Illinois at Urbana- Champaign.

Steve Cooper is president and CEO of One Stop Systems, and has more than 18 years’ sales, marketing, and general management experience in the standard bus-board marketplace. He began his career with Intel, where he became a technical spokesman for the concept of board-level open-bus standards and the Multibus and Multibus II architectures. Steve then joined RadiSys, a company specializing in embedded PC-compatible computers. He also served as vice president of sales and marketing, and later president and chief operating officer, at I-Bus. Most recently, Steve was president and chief operating officer for SBS Technologies. He holds a BSEE degree from the University of California, Santa Barbara.

For further information, contact Chris at:

PLX Technology, Inc. 870 Maude Ave. Sunnyvale, CA 94085 Tel: 800-759-3735 Fax: 408-774-2169 E-mail: [email protected] Website: www.plxtech.com

For further information, contact Steve at:

One Stop Systems 735 South Vinewood St. Escondido, CA 92029 Tel: 760-745-9883 Fax: 760-745-9824 E-mail: [email protected] Website: www.onestopsystems.com

CompactPCIRSC# 17 @www.compactpci-systems.com/rsc and AdvancedTCA Systems / November 2004 / 17 SOFTWARESOFTWARE CORNERCORNER Sun Microsystems announces By Curt Schwaderer AdvancedTCA plans at ITU CompactPCI & Telecom Asia AdvancedTCA

During the ITU Telecom Asia 2004 con- the technology but were having problems AdvancedTCA systems? The JES is well ference, Sun Microsystems announced combining complete AdvancedTCA based packaged with open interfaces enabling plans to launch an AdvancedTCA solutions effectively. Sun plans to lever- software on third-party blades to inte- compliant product family supporting age its key technologies, which include grate into the overall JES environment. both the Solaris OS and Carrier Grade CPU, OS, Java, and system management While Sun did not have any comment Linux operating systems. Sun also and high availability software into com- on third-party partnerships or developer announced an AC powered version of plete AdvancedTCA managed systems ecosystems for their AdvancedTCA its UltraSPARC processor and popular that network equipment manufacturers environment, they did mention that they Solaris OS based Netra 440 system. The can integrate quickly, and network service would be ready to offer fully integrated Netra 440 is a four-processor Symmetric providers can operate efficiently. solutions to network equipment manu- Multiprocessing (SMP) system used pri- facturers through a variety of means. marily for operation and maintenance While Sun continues to advance the Complete product solutions including applications such as service control Solaris OS and continues to optimize its AdvancedTCA chassis, switch cards, points, home location registers, call operation on UltraSPARC, the company software, and in the future, network pro- detail, and billing applications. The is also making Carrier Grade Linux avail- cessor unit blades for high-speed data product line targets network equipment able on the platforms as well. This allows plane processing could find their way manufacturers and telecom deploy- Sun to provide a complete, interoperable into the Sun AdvancedTCA solution set. ments but could find applications where system that may have a mixed Linux/ symmetric multiprocessing systems Solaris environment within a shelf. In In addition to the AdvancedTCA product are needed. turn, network equipment providers gain plans announcement, Sun also announced the flexibility of using Linux and Solaris that Telecom New Zealand has commit- Sun’s AdvancedTCA plans announce- based applications within their systems. ted to Sun’s enterprise server solutions ment is significant in terms of support for that include the Sun Fire 25K, Solaris 10, AdvancedTCA based systems in the mar- Sun’s Netra 440 systems are among the best and JES. ketplace. In a follow-up conversation with price/performance four-processor SMP Pocheng Shyu and Sandeep Agrawal, prod- systems on the market today. Standard From last summer’s SUPERCOMM uct managers for Sun Microsystems, they Performance Evaluation Corporation AdvancedTCA Pavilion to the Sun Micro- see strong demand from network ser- publishes benchmark information that systems announcement for AdvancedTCA vice providers for AdvancedTCA based includes the Netra 440 system. For more based enterprise servers, it appears that equipment. Sun’s announcement to put information, visit www.spec.org. AdvancedTCA is building momentum for UltraSPARC and Solaris technologies real systems and deployed solutions in in an AdvancedTCA form factor creates Sun’s network management software years to come. a bridge for network service providers package consists of the Java Enterprise already using these technologies to en- System (JES). The three main compo- For further information, contact Curt by hance their systems and services using an nents of the JES are: e-mail at cschwaderer@opensystems- industry standard form factor. Network publishing.com. service providers seem to be providing ■ Identity server significant pull for AdvancedTCA sys- ■ Web server tems, which promise higher performance, ■ Directing server higher levels of reliability and maintainability, cost efficiency, and a higher level However, there are a number of applica- Companies of flexibility than custom form factor sys- tions in the package that reduce the cost mentioned in tems provide. This trend also bodes well and complexity of buying, configuring, and this column: for the future of AdvancedTCA. integrating enterprise management, work group, and business-to-business software Sun Microsystems As a part of the AdvancedTCA effort, suites. For more information, visit www. www.sun.com/software/ Sun cites long lead times to combine sun.com/software/javaenterprisesystem/. javaenterprisesystem/ AdvancedTCA based systems that include blades, chassis, switch cards, and Certainly with the JES, Java, and Solaris Standard Performance system management software. This seems software environments for their planned Evaluation Corporation to confirm what I heard at last summer’s AdvancedTCA systems, the question www.spec.org SUPERCOMM from network service is: How do third-party blade and switch providers who seemed very interested in manufacturers integrate with Sun’s

18 / CompactPCI and AdvancedTCA Systems / November 2004 CompactPCI and AdvancedTCA Systems / November 2004 / 19 RSC# 19 @www.compactpci-systems.com/rsc STARFABRICSTARFABRIC WATCHWATCH

By Pat Hart StarFabric picks up the pace CompactPCI & AdvancedTCA

With the resurgence in the worldwide economy, the pace of in-order data streams and guaranteed bandwidth. StarFabric has StarFabric design wins and shipments has accelerated. The excelled in advanced systems where multiple hosts are required amount of coverage in the engineering press dedicated to to handle the complex acquisition, processing, and display tasks advances in serial interconnects and the emergence of PCI in digital imaging systems. These StarFabric features and this Express technology will no doubt help this positive trend level of performance have led to a number of design wins in this continue. In this column Pat gives an overview on several of space. the application areas that have contributed to the continuing StarFabric success. Advantest, a leader in the automated test of semiconductor components, recently revealed that several of its latest automated test While PCI Express has targeted early adoption in PCs and related equipment machines are using StarFabric as the interconnect peripherals, Switched PCI is the concept of passing PCI traffic between the central processor and the remote test head contain- through a point-to-point, switched serial interconnect to break ing the electronics that interface to the unit under test. The high- through the limitations in scalability of connections and physical speed serial nature of StarFabric and its ability to span rack to distance presented by the parallel bus approach. Since PCI has room scale distances allow this test head to be at a considerable been so widely adopted in all classes of embedded computing distance from the central computer. This arrangement increases and communication hardware, there is a great need to extend this configuration flexibility and the ability to interface with a vari- investment in software and legacy hardware to meet the increas- ety of automated handling equipment. Additionally, Advantest ing demands of tomorrow’s systems. Using StarFabric to extend engineers appreciated the smooth PCI-StarFabric integration and the size and reach of the tree topology of a PCI bus or bridged were able to shorten development cycles accordingly, while sig- PCI bus is a well-designed solution that requires no changes to nificantly improving the design process. application software, OS, or device drivers. StarFabric, in address routing mode, will be totally transparent to any OS with support In another recent announcement, CONTEC, an international for PCI bridging. In addition, the PCI bus scan recognizes all con- industrial computer and automation leader, announced a line nected devices at system start-up. These additional buses appear of expansion PCI chassis with a built-in StarFabric interface, as bridged PCI devices and can be on the same board, in the CardBus, and PCI StarFabric cards to expand PCI buses from a same chassis, or cabled to the upstream bus through up to 11 m notebook or other computer. Figure 1 shows the CONTEC PCcard of inexpensive CAT5e cable. The low cost and extended distance StarFabric adapter and PCI Expansion chassis. CONTEC chose offered by this cabling solution will not be possible with PCI StarFabric for its new extended adapter for bus extension system Express, thus providing a key advantage to StarFabric in PCI solution because of its 2.5 Gbps serial interconnect, reconfigu- expansion applications. The simple plug-and-play approach with rable fabric features, 100 percent backward compatibility to PCI, StarFabric is adequate for many applications, but the capabilities and ability to enable an external interconnect up to 40 feet over of the technology are far greater. CAT5e cable. The use of the CardBus form factors is unique and ideal for industrial and instrumentation applications and for pro- A Software Development Kit (SDK), and example code from viding a complete PCI expansion solution. Interest in this solu- StarGen accelerate the process of employing advanced StarFabric tion has been very encouraging. features, such as path routing and multiple classes of service. These advanced StarFabric features are beyond the scope of the Industrial computing has not been the only active market for device set up for PCI devices and are accomplished through the StarFabric designs. Military electronics suppliers have made great StarFabric-aware device driver software. Drivers for Windows use of StarFabric in a number of different applications. In previ- 2000 and XP as well as Linux are available from StarGen as part ous issues, this column has detailed StarFabric products from of the Evaluation Board Kit products. Source code for Windows, Linux, and VxWorks are available in the StarFabric SDK, and a driver for the Green Hills Software INTEGRITY RTOS has recently been added to the StarGen product set. Solflower Computer is supporting the Solaris market with hardware and device drivers. Using the SDK, an engineer can develop a driver for any operating environment.

For critical functions, quality of service and failover redundancy are necessary to ensure real-time results and guaranteed uptime. Many of the designs using these features are in military applications for the obvious reasons, but more industrial and commercial applications are requiring extended performance, reliability, and availability. Such applications include enterprise networking and storage systems as well as video systems that require Figure 1

20 / CompactPCI and AdvancedTCA Systems / November 2004 Dy 4 Systems, now part of Curtiss-Wright Controls Embedded Computing, Synergy Microsystems, also now part of Curtiss- Wright Controls Embedded Computing, and Innovative Integration. Texas Memory Systems has been the most recent military electronics manufacturer to announce a StarFabric product. It manufactures high-performance Digital Signal Processors and solid state disk storage systems. The SAM-650 DSP Supercomputer boasts 192 GFLOPS of processing power and has been upgraded to include StarFabric I/O. With this increased performance, the SAM-650 DSP eliminates most DSP performance bottlenecks that are often associated with processing high bandwidth signals. It was designed specifically for a government agency customer and will likely be designed into multiple, military- grade embedded systems requiring extremely high performance and reliability. Micro Memory has also announced VME carrier cards with two StarFabric ports and two 64 bit/66 MHz PMC sites. Figure 2 is the MM-6450D Othello real-time data acquisition and signal processing system board from Micro Memory. Othello rates buffer data between I/O streams and StarFabric.

More than 75 companies now use StarFabric in their products. Star- Fabric port shipments are at 250,000, and recent de- Figure 2 sign wins suggest the run rate is accelerating. With many of these designs now moving to volume production, shipments will ramp up for a number of years. Recent design wins will start generating significant vol- umes later in the year and in 2005, and many of these products have very long lives.

As new serial interconnect technologies, such as PCI Express and Advanced Switching Interconnect are introduced, StarFabric offers embedded OEMs the benefits today that the emerging switch technologies promise in the future. As a cost-effective, expandable switched PCI solution, StarFabric is a suitable interconnect technology for embedded computing applications, including industrial control, semiconductor manufacturing and automated testing, medical imaging, video distribution, and a range of high-performance military applications.

Pat Hart manages marketing communications at StarGen. Prior to StarGen, Pat was with API Networks, Compaq Computer, Digital Equipment Corp., and Texas Instruments. He has 20 years’ experience in marketing and design in the electronics and computer industry. He holds a Bachelors degree in Electrical Engineering from Marquette University.

For further information, contact Pat at:

StarGen, Inc. 225 Cedar Hill Street Suite 22 Marlborough, MA 01752 Tel: 508-786-9950 Fax: 508-786-9785 E-mail: [email protected] Website: www.stargen.com RSC# 21 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 21 SPECIAL F E A T U R E :

EMBEDDED SOFTWARE Manageability in AdvancedTCA and the role of IPMI

By Sanjoy Maity

n this article Sanjoy discusses the Management architecture of ■ An IPMI infrastructure that provides use of graphical tools for sensor AdvancedTCA communications, management, and layout design when porting IPMI As established by the PICMG industry control between the distributed man- Ifirmware onto different platforms. organization, the AdvancedTCA speci- agement controllers and to an overall fication provides a strong foundation for System Manager. A mirrored board The IPMI and AdvancedTCA specifica- system management. The management provides central management through tions are now well known, but still rep- model is a multi-tier architecture. At the the IPMC. resent a relatively new technology that highest level, the System Manager is ■ A higher level, high-speed TCP/IP is slowly replacing older proprietary responsible for managing one or more service for boards that need remote schemes. The lengthy porting proce- shelves and possibly one or more systems. booting, SNMP management, remote dure involved with incorporating the The System Manager watches hardware disk services, or other management IPMI firmware onto different platforms health and other management events of operations. with different sensor configurations can the multiple shelves. On the second level, be remarkably shortened by the use of there are two Shelf Managers in active- AdvancedTCA adds further crucial ele- graphical tools that enable engineers passive configuration. The main responsi- ments to the IPMI management structure to simply drag and drop a sensor from bility of these Shelf Managers is to moni- such as hot swapping of boards, power a list to the device connection window. tor and manage the low-level entities. supplies, and other FRUs, redundant As technology developers our desire is IPMB buses for management traffic, and always to have longer technology life- IPMI: Its contribution to E-Keying for checking the backplane’s cycles for the products. Open standards AdvancedTCA interface compatibility before enabling it. offer such ability and allow us not to The blade units are considered Field depend on a sole technology provider. Replaceable Units (FRUs) in Advanced- IPMI: Specification This brings tremendous competition TCA systems. All the units composing IPMI provides the backbone (basic) man- and opportunities to grow global inno- an AdvancedTCA system are required agement infrastructure for AdvancedTCA vations in the industry. AdvancedTCA to comply with the IPMI specification, systems. Each board must contain a man- (Advanced Telecom Computing Archi- now in version 2.0, by implementing agement chip, defined as Baseboard tecture) evolved as an outcome of such an Intelligent Platform Management Management Controller (BMC) under the open architecture momentum. Industry Controller (IPMC). The main manage- IPMI 2.0 specification or IPMC under the standards also offer common ways to ment bus within the shelf is also part of AdvancedTCA specification. interface among different modules from the IPMI architecture, consisting of two different vendors in both hardware and Intelligent Platform Management Buses The IPMI model is based on a specialized software aspects. (IPMBs). The communication between the microprocessor, the BMC, which handles shelf manager and other IPMCs should be health monitoring, alert, and management In 2001 as a result of this open stan- performed in a round robin fashion with functions independently of the main proces- dard initiative, PCI Industrial Computer commands alternating on the two IPMBs. sor. Major system modules are connected Manufacturers Group (PICMG) started to the BMC by the redundant IPMB. formulating a standard that could be used AdvancedTCA provides manageability by the telecom industry for advanced and redundancy at the shelf level, while The BMC operates on standby power computing. The specification has gone IPMI describes standard management and autonomously polls system health through several evolutions, giving birth structures at the board level. All PICMG variables such as temperature, fans, volt- to today’s AdvancedTCA solution. specifications are based on the Intelligent age, power supplies, and chassis intru- AdvancedTCA is the industry standard Platform Management Interface (IPMI) sion. When discovering any anomaly, way to develop computing blades or com- management standard. AdvancedTCA System Management Software (SMS) ponents that satisfy the telecom industry’s robust management and availability fea- can log the event through the BMC, send reliability, availability, and manageabil- tures are based on three elements: alerts, and initiate recovery actions such ity demands. AdvancedTCA provides as system reset. low-level hardware based management ■ Distributed management controllers components to the high-level software to monitor the operation of each To operate successfully, the IPMI model management entity. system FRU. requires a software management layer

22 / CompactPCI and AdvancedTCA Systems / November 2004 running under the operating system (Figure 1). The SMS has in-band access to the IPMI information and commands, providing enhanced manageability and threshold/event management.

Access to the monitoring hardware is not provided directly, but through abstracted sensor commands implemented by the BMC. In other words, IPMI defines a sensor model using message- based interfaces. The various event and sensor types are represented by numeric codes provided by the IPMI specification. The BMC provides a centralized System Event Log (SEL). A specific set of IPMI commands registers events on the SEL, which is available for reading and analy- Figure 1 sis even if the system is totally nonfunc- tional and inaccessible.

The SMS retrieves platform information and automatically configures itself. Platform information is provided by Capabilities commands and Sensor Data Records (SDRs). Capabilities commands provide information fields on various fields and commands under control. SDRs feed data to the SMS about the type and number of sensors in the platform, sensor threshold support, event generation capabilities, and readings information. The BMC acts as a sensor initialization agent.

A centralized SDR repository located on nonvolatile storage provides SDR information via out-of-band interfaces like the Intelligent Chassis Management Bus or via a remote access card.

IPMI requires that information on FRUs be stored and accessible via SMS also when the main processor becomes unavailable. FRU information typically includes serial number, part number, model, and asset tag.

Providing access to vital data also out-of- band or under system down conditions, the IPMI FRU inventory is complemen- tary to information provided by SMBIOS and PCI Vital Product Data. IPMI supports three types of system interfaces:

■ Keyboard Controller Style (KCS) ■ System Management Interface (SMIC) ■ Block Transfer (BT)

The BMC has communication capabilities via LAN, or sideband, by utilizing messages encapsulated in the Remote Management Control Protocol (RMCP) packet datagram. RSC# 23 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 23 SPECIAL F E A T U R E : EMBEDDED SOFTWARE

IPMI supports LAN alerting in the form of SNMP traps that follow the Platform Event Trap (PET) format.

The specification provides a standard interface for a system watchdog timer, a capability usually utilized by enterprise- class servers. A watchdog timer can be configured to automatically perform actions such as power off, power cycle, reset, and interrupt. (See Figure 2).

Improvements in the latest 2.0 specification Version 2.0 of the specification improves IPMI communication and authentication capability.

■ Serial over LAN – The BMC serial controller traffic can be redirected Figure 2 over the LAN, providing text redirection over a remote console. AMI had causing the CMM to shut down an An IPMI firmware implementation should already introduced this feature in its entire AdvancedTCA system. be able to provide full IPMI functionality IPMI 1.5 implementation. with minimal OS and platform depen- ■ Support for multiple payloads – An An AdvancedTCA management dency, to ease and speed up eventual port- IPMI 2.0 session can carry types of solution ing tasks. Design goals should include traffic that are additional to IPMI A superior IPMI stack and an minimal footprint, easy sensor configura- messages. AdvancedTCA management solution tion, and state-of-the-art security in order ■ Firmware Firewall – The Firmware are available from AMI for the telecom to guarantee short development time and Firewall provides a mechanism for industry. MegaRAC-PM, the IPMI stack, maximum reliability to AdvancedTCA the BMC to disable/enable certain keeps the focus on the configurabil- board designers and TEMs. commands from specified interfaces, ity, extendibility, and robustness of the such as the System Interface, and to IPMI 2.0 based IPMC solution for mak- Addressing implementation launch a Command Discovery pro- ing AdvancedTCA. The outcome of this challenges by using graphical cedure that lets the software discover effort offers a strong, intuitive graphi- tools for sensor layout design which commands and functions are cal tool to configure the sensors and its AMI IPMI firmware, MegaRAC-PM, available. The firewall is introduced management, referred to as Platform is built over an RTOS-based microker- primarily to hamper malicious code to Management Configuration Program, or nel, which simplifies porting to differ- simulate a nonexistent system failure, PMCP, hereafter. ent microcontrollers. OS and platform

RSC# 24 @www.compactpci-systems.com/rsc

24 / CompactPCI and AdvancedTCA Systems / November 2004 dependency are kept to a minimum. The schemes. The lengthy porting procedure firmware is written in platform indepen- involved with incorporating the IPMI dent code. The platform specific features firmware onto different platforms with are abstracted as macros; higher software different sensor configurations can be layers use only these macros to access remarkably shortened by the use of graph- the hardware or OS functions. All that is ical tools that enable engineers to simply needed for porting to different platforms drag and drop a sensor from a list to the is to rewrite these specified macros and device connection window (Figure 3). their associated functions. Firmware foot- Such tools should provide a way for print is kept to a minimum, also thanks to OEMs to specify how each GPIO is used, a table-driven structure, resulting in com- how each is connected to the BMC on a pact and efficient code. specific platform, and whether each one is interrupt-driven or not. An example An IPMI implementation will definitely can be AMI’s PMCP, integrated with the benefit from interaction with the system company’s IPMI firmware. BIOS, although a design without any BIOS support is theoretically possible within Sanjoy Maity is vice-president of soft- the boundaries of the IPMI specification. ware engineering, American Megatrends AMI’s BIOS module, for example, adds Inc. (AMI) He holds a BE in Computer several management advantages to a sys- Science and Engineering from Calcutta tem equipped with AMI IPMI firmware: University, India. He has 18 years of work experience in embedded firmware ■ POST code logging into the BMC and software development in enterprise System Event Log server management products. He has ■ Memory errors logged into the SEL experience in various engineering posi- ■ Non-recoverable errors logged into tions at American Megatrends. Sanjoy the SEL has supervised the development of AMI’s ■ Watchdog timer interaction (set/reset) remote control cards, Network Attached ■ Synchronize SEL timestamp with the Storage, and IP-SAN products. system Real-Time Clock (RTC) for accurate event log information For more information, contact Sanjoy at: ■ Text redirection through serial port and LAN American Megatrends Inc. ■ Boot timeout management 6145-F Northbelt Parkway Norcross, GA 30071-2976 The IPMI and AdvancedTCA specifica- Tel: 800-828-9264 tions are now well known, but still rep- Fax: 404-263-9381 resent a relatively new technology that E-mail: [email protected] is slowly replacing older proprietary Website: www.ami.com

Figure 3

RSC# 25 @www.compactpci-systems.com/rsc SPECIAL FEATURE:

EMBEDDED SOFTWARE The role of standards based high availability middleware in accelerating product development

By John Fryer

igh availability has been an essential requirement encompassing a wide range of suppliers. However, selecting a for mission critical applications for many years. The set of perceived best-in-class blade level solutions, along with H telecommunications industry exemplifies this criterion chassis, switching, and platform control processing, still presents with the concept of 5-nines and 6-nines uptime. Traditionally, substantial development challenges. high availability has been a proprietary development, often linked closely to product architecture and the belief that this Many embedded technology applications operate in environ- functionality delivers market differentiation. Today, the econom- ments where high availability is a critical requirement, with ics of many industry segments are changing. The business fac- unscheduled downtime being measured in minutes or seconds per tors that are becoming increasingly dominant over proprietary year. The high availability software systems that provide these developments include: Time-to-market, cost efficiency through mechanisms have traditionally been in-house developments that streamlined development and supply chain processes and invest- have evolved over many years. They represent hundreds of man- ment protection, and leveraging standards based technologies. years of development and testing, with 20-30 man-years of support effort required per year. Although it has often been viewed as AdvancedTCA and CompactPCI are primary examples of market differentiating by technology organizations, proprietary standards based hardware. When complemented with stan- middleware development and support places an increasing bur- dards based high availability middleware, based on Service den on many companies. In markets driven by time-to-market, Availability Forum (SAF) standards to create Application cost, and lastly, differentiation, nuances in high availability fea- Enabling Platforms, the combination offers even more ben- tures do not support the investments required to sustain them. The efits in achieving product flexibility and industry desired price requirement to support 5-nines or 6-nines is the price of entry to points. In this article John explores both the advantages of high availability market segments. These features themselves are off-the-shelf high availability middleware and the challenges simply base requirements for the users of that equipment. It is faced by industries, such as telecommunications, to maximize the applications on top of high availability that drive revenue and the potential such products can offer. differentiation. This in no way detracts from the creative effort applied to understand and implement the services required for The economics of many technology segments have changed dra- high availability. The challenge is to leverage this work into a matically in the last several years, no more so than in the tele- standards effort that streamlines developments in an equivalent communications sector. This is an industry once dominated by manner to those provided by AdvancedTCA and CompactPCI. technological advances and high system prices supported by a customer base of service providers operating substantially within The Service Availability Forum is pioneering the standardization a monopoly. Deregulation has resulted in transition to a highly of middleware components necessary for 5-nines and 6-nines ser- competitive environment, as evidenced by continually falling vice availability. (See Figure 1). Much of this work is leveraged voice services pricing. The flow down pressure on equipment from the expertise developed by companies who have developed vendors has led to streamlined operations and smaller develop- their own proprietary middleware. The standards are framed in ment teams, yet the necessity to bring increasingly sophisticated products to market in a shorter time has only intensified. Software applications primarily represent the differentiating value these teams bring to the product creation process.

The AdvancedTCA specifications from PICMG provide the first large scale form factor, power profile, and thermal and mechanical characteristics that are suitable for products targeted for applications such as telecommunications’ network cores. For many years, CompactPCI has been used in an ancillary role where cost and rapid product development have been key factors. The use of these two technology families is now firmly established as they offer a broad ecosystem Figure 1

26 / CompactPCI and AdvancedTCA Systems / November 2004 CompactPCI and AdvancedTCA Systems / November 2004 / 27 RSC# 27 @www.compactpci-systems.com/rsc SPECIAL F E A T U R E : EMBEDDED SOFTWARE a set of Application Programming Interfaces (APIs) that define ■ RNC functionality for 2.5 and 3G wireless networks the implementation parameters for the set of services required to ■ Traffic engineering applications for IP networks implement service availability. The currently available standards ■ Integrating IP routing focus on a Hardware Platform Interface (HPI) that provides infor- ■ Multi-Protocol Label Switching (MPLS) mation on basic platform characteristics such as fan speeds and ■ Quality of Service path calculation engines blade status. This can be used to make failover decisions caused by basic hardware failure. The Application Interface Specification Once applications migrate to using the AIS APIs it becomes much (AIS) provides APIs for the range of services required to perform easier to port these applications across multiple platforms. Clearly, service availability at the application, primarily software layer. the underlying architecture, processing power, and memory may This includes such capabilities as: differ among platforms, but the use of standard APIs means that it is inherently easier, and faster, to complete development. ■ Cluster membership (system description) ■ Message service Indeed, so powerful are the SAF concepts and specification ■ Locking service that related standards organizations are adopting them at their ■ Check-pointing service boundary layers. The Open Communications Applications ■ Event distribution service Forum (OCAF) is adopting SAF standards in its efforts to define a standardized architecture for communications equipment. The next iteration of the SAF standards will refine the current work Additionally, the Network Processor Forum (NPF) is adopting and add more services along with a systems management layer for SAF APIs as its standard interface for service availability func- comprehensive set of standards that will cover most applications. tionality, and the Open Source Development Lab (OSDL) is working closely with SAF to ensure that Linux aligns with the The benefits of SAF extend well beyond the concept of standards HPI and AIS specifications. based middleware. A standard set of APIs for implementing service availability constructs a bottom layer interface for platform Standards and off-the-shelf implementations to those standards applications. In the telecommunications world these platform offer some measure of acceleration of the development process. applications include capabilities such as protocol implementa- However, they still leave considerable work for development tions and integrated protocol applications. Examples include: teams before real product differentiation can be achieved.

RSC# 2801 @www.compactpci-systems.com/rsc RSC# 2802 @www.compactpci-systems.com/rsc

28 / CompactPCI and AdvancedTCA Systems / November 2004 Application Enabling Platforms pre-integrate a combination 2005. As applications built to SAF APIs also begin to appear in of PICMG and SAF standards plus the Linux operating system, the marketplace, the broader ecosystem encompassing multiple offering a whole new dimension in capability. (See Figure 2). standards is taking shape. Rather than threatening development This opportunity is recognized by the increasing liaison between organizations, the Application Enabling Platform approach offers PICMG and the Service Availability Forum. These platforms compelling advantages to enable the technology and business offer competitive advantages, cost efficiencies, and investment arms of equipment vendors to meet the market pressures of cre- protection in a variety of ways. Competitive advantages are ating competitive advantages, accelerating time-to-market, and realized through time-to-market acceleration from utilizing pre- protecting investments. integrated and verified hardware and software platforms. This enables resources to be focused on product differentiation and John Fryer is the director for worldwide marketing of Motorola also reduces efforts in standards compliance and system testing. Embedded Communications Computing Group for Advanced Reusing off-the-shelf platforms across multiple applications gen- Platforms, where he focuses on assessing and developing new erates cost efficiencies. Standards also increase cross-platform market and product opportunities for the company and assuring synergy and facilitate backward and forward compatibility. This that the needs of its customers – telecommunications equip- latter point is also important as it provides investment protection ment manufacturers – are being met. John became part of the through successive generations of hardware and software. Motorola team after NETPLANE Systems, Inc. was acquired by Motorola in early 2003. John represents Motorola at stan- The overall risk reduction that is offered by the Application Enabling dards bodies and on industry panels, leveraging his more than Platform approach is particularly significant. With the guaranteed 24 years of experience in the communications industry. He led interoperability offered by standards based hardware and software international marketing initiatives at NETPLANE Systems, Inc. components, users of the technology have the freedom to select the John holds a Bachelors degree with honors from the University best-in-class products that meet their own criteria. of Nottingham, Great Britain.

Since Application Enabling Platforms are based on AdvancedTCA, For further information, contact John at: CompactPCI, OSDL compliant Linux, and SAF standards they offer a choice of integration points. Users can adopt any level of Motorola Embedded Communications Computing Group these standards based on their own decisions concerning where 2900 S. Diablo Way DW205 and now they provide value in the marketplace. For some com- Tempe, AZ 85282 panies, this means reliance on standards based hardware, with Tel: 602-437-6701 • Fax: 602-437-6701 continued use of in-house middleware and existing applications. E-mail: [email protected] Increasingly, the time-to-market factors and cost efficiencies Website: www.motorola.com/computers offered by enabling applications to easily operate across a range of platforms is providing the impetus to migrate to pre-integrated Service Availability Forum standards.

In addition to the technical advantages of adopting Application Enabling Platforms, the streamlining of the value chain offers significant options for adopters. This provides the choice to source products in traditional ways or to utilize delivery of pre-configured and tested Application Enabling Platforms, even shipping these directly to end users.

This vision of standards based Application Enabling Platforms, where both hardware and a range of software layers are pre- integrated, is taking hold in the marketplace. The economics of migrating to AdvancedTCA and CompactPCI have already been established. SAF compliant products are starting to appear in the marketplace, and multiple implementations can be expected in

Figure 2 RSC# 29 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 29 SPECIAL F E A T U R E :

EMBEDDED SOFTWARE Best practices for secure console port management: Establish a virtual crash cart to minimize downtime every time

By Charles Pheterson

dvancements in technology continue to enable busi- devices, surge protectors, and intelligent power strips that keep nesses to better serve their customers. In some areas, them all powered. A notably supply chain management, order entry, and inventory management, these advancements have enabled orga- Most of these devices support two means of access: user access nizations to achieve extremely high levels of efficiency. Products and maintenance access. Servers, for example, when function- are being developed, manufactured, customized, and delivered ing normally, are accessed by users via Web clients, SQL cli- faster than ever, supported by back office automation. Data ents, e-mail clients, and FTP clients. User access relies upon the centers facilitate automation, housing business-critical servers proper functioning of the server, the operating system, the appli- that communicate across a mix of private and public networks. cation, the protocol stacks, and indeed the network infrastructure Reliance upon centralized services creates a new level of risk between the client and the server. When failures occur, user level for the businesses that depend upon them. Data centers are a interfaces are often nonresponsive, and so cannot be relied upon fulcrum upon which the fate of an automated business is precar- to restore a failed device back to normal operation. iously balanced. When fully operational, business hums along at speeds not previously possible. When downtime strikes, however, IT staff need a backdoor interface to maintain reliable mainte- business screeches to a crashing halt. nance access to data center devices (Figure 1). The backdoor has to function independently from the operating system, applica- Data centers form the heart of business automation systems. tions, protocol stacks, and production networks because if any When failure occurs in a data center, business bears the impact. one or more of those is down, access is crucial to repair. The As such, an IT manager’s number one imperative is to maintain serial console port, which is fairly universal, supplies this back- uptime. To manage downtime, it is crucial that administrators be door interface. Requiring only BIOS-level functionality from the organized, respond quickly, manage users, tighten security, and device, it provides critical access via a dumb terminal, so that act proactively. With IT staffing often stretched thin across dis- IT staff can run diagnostics, reconfigure corrupted settings, view tributed data centers, it is difficult to be there to provide hands- status, and restore failed units to normal operational status. on maintenance and problem resolution; data center staff need to have virtual crash cart access to business critical devices. In Backdoor access for repair, monitoring, and this article Charles details how console management solutions maintenance can provide virtual crash cart access, thus offering a means In UNIX and Linux servers, the console port serves as a privi- by which administrators manage devices and respond to crises leged interface even during normal operation. Critical kernel quickly, securely, and proactively. and application messages are often directed to the console port as a form of event log. Server management of non-VGA servers The challenges of a heterogeneous data center is often performed via the system console. With the advent of Today’s data centers contain a wide range of devices including:

■ Application servers ■ E-mail servers ■ Database servers ■ Web servers ■ E-commerce servers

These servers are often a mix of Intel- and RISC-based with legacy minicomputers and mainframes. Operating systems vary, too, with many IT groups maintaining a variety of Windows, UNIX, Linux, Mac OS, and/or Netware systems. Beyond the servers themselves, data center staff must also support the SAN and NAS systems used for shared storage, the collection of networking equipment to make it all communicate, and the UPS Figure 1

30 / CompactPCI and AdvancedTCA Systems / November 2004 Windows 2003, the serial console now also provides direct access to the Windows Emergency Management Services. These services operate at a nonkernel level and offer the ability to restore normal functionality even in the event of blue-screen.

The dumb terminal is the most basic means of out-of-band access to failed devices. Most data centers have a crash cart dumb terminal that is wheeled about for emergency management of failed systems. Direct terminals bypass protocol stacks, operating systems, routers, and switches. Unfortunately, administrators are not always in the data center when a failure occurs. Factors such as downsizing, business travel, multi-site facilities, branch offices, lights-out policies, and 24x7 operations continue to increase the likelihood that failures will occur when administrators are not present. Figure 3 Because administrative staff are often absent when trouble strikes, organizations have turned to console management solutions as a key building block in their strategy to minimize down- they provide a truly comprehensive console management solution time. Whether local or remote, console management solutions with speed-dial access to data center devices via a database-driven give administrators secure access to console ports of data center application that serves as a desktop for the entire data center. devices. These solutions ease the burden by providing virtual crash cart access to every device in the data center. Console man- Getting specific agement also affords administrators a host of valuable tools to AVWorks software brings administrators a desktop view of the detect and diagnose failed units as well as to quickly restore them entire data center. Its user interface makes it easy to identify to fully operational status. and access servers, routers, and operating systems via icons and clearly visible device names. In addition, IT personnel can enter The office Rolodex is a familiar concept to anyone with a large additional properties for each device – type, location, and other number of contacts to manage. The feature is a response to the asset management information, or even administrators’ names need for fingertip access to critical information. The Rolodex and phone numbers or comments – that will display with a right- allows you to quickly access contact information that is most click on the device icon. important to you. Data center devices also require contact information:

■ IP addresses ■ System names ■ Administrator names ■ Contact information

Where is this data stored? In many cases, it is relegated to memory and post-it notes. When trouble strikes, action must be taken quickly. Fingertip access to device names and IP addresses is a must.

Applications such as the Avocent AVWorks cross-platform management application, shown in Figure 2, when combined with a console management appliance like the Equinox CCM Console Manager (Figure 3), make it easier to access devices and the information needed to remedy problems with them. Together,

Figure 2 RSC# 31 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 31 SPECIAL F E A T U R E : EMBEDDED SOFTWARE

The AVWorks software search, sort, and grouping features ease ■ The appliance’s ability to act as a PPP server the task of locating failed units. Sort bars allow devices to be ■ Appliance support for authentication and access control on sorted by device type, site, department, or location. For example, dial connections an administrator could look at just Cisco routers or Linux servers. Group folders narrow the display to a particular device type, site, Look, too, for advanced features such as IP forwarding to facili- or location. Search makes it easy to locate one or more pieces of tate the sharing of vital modems. IP traffic arriving at the serial equipment via custom search capability that allows administra- port via the modem connection is redirected out the network port tors to search any field. So, a search for Joe Smith, for example, of the appliance, thereby making it possible to access the console could display all of the devices for which Joe is responsible, or port of any device in the data center to restore functionality to a search for Data Center B could display all of the equipment in failed devices. Data Center B. Security counts Wizards make adding items to the AVWorks management soft- Network administrators and server administrators need access ware database easy. The software automatically discovers CCM to different devices. Likewise, administrators of Windows/Intel Console Manager appliances, device names, and even third-party servers require access to a different set of systems than UNIX/ appliances and servers that respond to SNMP queries. In addition RISC administrators. To enforce access policies within the data to reverse Telnet and SSH sessions, AVWorks management soft- center, it’s important to select a console management solution that ware allows in-band Telnet and HTTP sessions to be launched to offers robust options for user authentication and access control. all target devices, including third-party appliances and servers. A Devices tab presents a list of devices that are attached to console Most appliances on the market provide an option to create internal management appliances, including the Equinox CCM appliance user accounts. The CCM Console Manager, for example, contains and Avocent KVM over IP switches. An Appliances tab presents an onboard database for creating 64 user accounts. Specific access a list of console management appliances and third-party devices. rights may be set for each account. For example, an account cre- This tab is useful when adding new appliances or configuring ated for Joe Smith may grant him access only to devices attached to existing appliances. CCM appliance ports 1, 5, and 9. Advanced features such as access control levels ease the task of assigning user access rights. With Easy access access levels, it is a simple matter to create new accounts and assign A click on a desktop icon within AVWorks management soft- permission. By way of example, the CCM Console Manager offers ware launches a console session to the selected device. If the three levels of access control: CCM Console Manager is configured for Telnet access, then the console session is launched in plain text mode. If the CCM ■ Appliance Administrator appliance is configured for SSH access, then a secure SSH ses- ■ User Administrator sion launches. ■ User

To further automate console access, AVWorks management soft- The CCM Console Manager appliance is factory-set to contain ware maintains specific connection parameters for each target one account: Admin, which has the broadest set of privileges, device in its database. For example, one server may require VT100 Appliance Administrator. This account is used to configure the emulation while another may require VT320. Likewise, separate appliance and create subsequent accounts. User Administrators log files, if desired, can be defined for management sessions with are key people responsible for systems attached to the console each device. When using AVWorks software, the mechanics of manager, but are not responsible for the console management console access are hidden from the user. While they’re easy to appliance itself. Accounts with User-level privilege are able to view or edit when necessary, IP addresses, port numbers, emula- access only those devices to which they are granted permission. tion modes, and other parameters need not be remembered once Accounts with User Adminstrator access are able to: they are stored into the application’s database. ■ Establish new user accounts AVWorks software also has a credential caching feature to ■ Define access rights streamline the sign-on process. With credential caching, users ■ Access devices attached to all ports of AVWorks software need only enter credentials once and, although each serial session is authenticated separately, creden- Internal user accounts are stored within the console management tials are stored and used to launch subsequent sessions. If cached appliance. Although these accounts are secure and simple to cre- credentials fail authentication, users are required to re-enter their ate, they may become cumbersome in organizations that maintain credentials. a great number of appliances. Consider an organization that has 15 CCM Console Manager appliances that are located at three Distance management – choose carefully different sites. Each time a user is hired or fired, it may be neces- Devices fail at inconvenient times and often when the adminis- sary to update the user database of each appliance. This process trator is not in the office. Remote access is a critical tool for fast may take several minutes per appliance. For those environments response, and the virtual device access afforded by a good con- with large numbers of console appliances or users, centralized sole management solution facilitates that quick action. Console user accounts are more appropriate and easier to administer. management must provide out-of-band access via dial-up connectivity. When selecting a console management solution, verifying A good console management appliance should also support a all of the following is important: means of external authentication, such as RADIUS. Originally

32 / CompactPCI and AdvancedTCA Systems / November 2004 developed for remote access dial-in user authentication, RADIUS to respond to system failures before they have an impact on the is now widely used for access to routers and other data center user community. SNMP traps help by notifying administrators of devices such as console management appliances. Authentication issues early so they can begin working on problems before users attempts that are fielded by the CCM Console Manager are passed even know they exist. to the external server via the RADIUS protocol. (See Figure 4). The server, in turn, provides access-right information back to the CCM Console management appliances, such as the CCM Console appliance if the user name and password are valid. If the account is Manager, typically include a SNMP agent. Via the agent, SNMP not valid, then the request will be denied. Security is best managed managers are able to view operational status and statistics. Some as a series of layers. Data encryption, although most certainly nec- appliances go a step further; they offer an enterprise MIB, which essary, is not sufficient on its own. Console management solutions includes unique value-added capabilities. Enterprise traps offer must provide security features across many dimensions. a real-time heads-up when a suspicious event occurs. The CCM Console Manager port alert feature allows administrators to specify ten unique text strings per serial port. The CCM appliance scans serial data, even when the port is not being accessed via a Telnet or SSH session. When a port alert string is detected, the appliance will forward the string to an SNMP manager via a trap message, as depicted in Figure 5.

Figure 4

Problem fixed – advanced features Administrators often become aware of failed systems via user complaints. By the time complaints occur, however, business has already been affected. Ideally, administrators should be able Figure 5

RSC# 33 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 33 Using the port alert feature, UNIX administrators are able to respond immediately to system console events that would have otherwise fallen upon deaf ears. The CCM appliance is also con- figurable to send a trap upon loss of control signals such as carrier detect from an attached device. Signal loss may indicate:

■ Failure of a power circuit ■ Failure of a power supply ■ Software anomalies ■ Cabling problems

Timely response allows failed units to be restored to operational status before the failure seriously harms business operations.

Conclusion System failures can result in huge losses for a business, and a great deal of user dissatisfaction. The need to manage downtime risk is greater than ever. To manage systems to minimize downtime, administrators must be organized, respond quickly, manage users and their expectations, tighten security, and deploy the right Figure 6 tools. Console management solutions like the Equinox CCM Console Manager with AVWorks management software provide administrators with exactly the tools to stay on top of systems and For further information, contact Charles at: keep things running smoothly (Figure 6). Equinox Charles Pheterson received a degree in computer science One Equinox Way from Nova University in Davie, Florida. His career started as a Sunrise, FL 33351 computer programmer and progressed to software engineering Tel: 800-275-3500 manager. He has served on several ANSI committees. Charles Fax: 954-746-9101 is currently a senior product marketing manager for Equinox, E-mail: [email protected] An Avocent® Company. Website: www.equinox.com

RSC# 34 @www.compactpci-systems.com/rsc

34 / CompactPCI and AdvancedTCA Systems / November 2004 APPLICATION F E A T U R E :

H O T S W A P CompactPCI hot swapping

By Alexey Polonsky

n this article Alexey describes swapping of other types like the Single 64-bit initialization. These chips CompactPCI board and silicon Board Computer (SBC). add the pre-charge and early power hot swap features and addresses support, so that no external circuitry the problem of mass-market oper- CompactPCI boards (other than the sys- is required, except for the onboard Iating systems’ lacking the additional tem host) can be one of three types: power control. software layers needed to support the new hot swap capabilities. ■ Non-hot swap: Boards that do not have Enumeration and dynamic all the features required for hot swap- configuration Introduction ping. They cannot be inserted or ex- The CompactPCI Hot Swap specification Hot swap is the ability to insert and remove tracted from an operating CompactPCI (PICMG 2.1 R1.0) defines enumeration boards, without adversely affecting a run- system. as the action taken by the system host to ning system. The CompactPCI Hot Swap ■ Basic hot swap: Boards with minimum poll the configuration spaces of all the specification allows vendors to add hot features required for hot swapping. PCI devices and assign or release the nec- swap capabilities to a CompactPCI based ■ Full hot swap: Boards that have the essary resources: system. minimum features plus additional resources for software connection con- ■ Memory and/or I/O address space The hot swap feature that has been estab- trol. Full hot swap boards allow the full ■ Interrupts lished in the CompactPCI specification is range of the system capabilities. ■ Software drivers particularly well suited, and has indeed been a critical requirement for applica- CompactPCI silicon can be one of four In systems that do not support hot swap, tion designs in the telecom and datacom types: enumeration only takes place as the markets. In systems that do not support system boots. In hot swap capable sys- hot swapping of peripheral boards, each ■ Non-hot swap: The silicon is not tems, dynamic configuration is required, board insertion or removal requires a compatible with the minimal require- whereby the system software allocates complete shutdown of the entire system ments of the CompactPCI Hot Swap (enumerates) system resources to the hot until the process is complete, in order to specification. swap board, following the insertion of the prevent damage to other boards or to the ■ Hot swap capable: Silicon that board. system software. In time-critical telecom contains the minimum requirements and datacom systems, system down- to operate in a hot swap environment. These same resources are released upon time is a major financial and quality of These minimum requirements are the extraction of the board. In order to service factor. Any downtime translates attributes that a board designer would take advantage of the hardware’s hot swap into financial loss and disconnecting ser- not be able to compensate for with capability, a substantial amount of addi- vice to active lines. Hence, reduction or external circuitry. tional software is required at the device even elimination of system downtime is ■ Hot swap friendly: Silicon that driver level, at the system services level, critical. contains support for software and at the applications level. connection control (as defined in The hot swap facility reduces the sys- PICMG 2.1 R1.0). Specifically, the The software aspect tem downtime, which may be required silicon supports a Control and Status The software aspect of the hot swap for replacement of boards and making Register in the PCI configuration process involves detection of the hot changes in the system configuration. The space and provides resources for swap event, dynamic configuration, hot swap feature is one of the key reasons ENUM#, hot swap switch, and the and resource allocation. It also involves why CompactPCI architecture is so attrac- blue LED. informing the system of the event so that tive for telecom and datacom vendors. ■ Hot swap ready: Silicon that con- the system and applications running are tains all the desirable features for not damaged, and the insertion/removal CompactPCI hot swap hot swap. All the desirable features event is transparent to the system user. Though CompactPCI hot swap is not for hot swap are all the features limited to peripheral boards at this time, supported by hot swap friendly While many hardware products are the CompactPCI Hot Swap specification silicon, in addition to Bias Voltage already electrically and mechanically hot (PICMG 2.1 R1.0) does not address hot Support, Early Power Support, and swap compatible, until now little software

CompactPCI and AdvancedTCA Systems / November 2004 / 35 APPLICATION F E A T U R E : H O T S W A P

existed that allowed utilization of the new ■ Identifying the insertion/removal of resources and (optional) dynamically capabilities. Mass-market operating sys- CompactPCI boards. allocate them or de-allocate them upon tems, such as Windows 98, Me, NT, 2000, ■ Notifying the system services and the removal of the board. XP, Server 2003, Solaris (until version 8), relevant drivers of a hot swap event. Linux, and VxWorks, currently lack the ■ Notifying the relevant applications of Kernel-Mode Messaging Mechanism additional software layers needed to sup- a hot swap event. This component is responsible for estab- port the new hot swap capabilities. The lishing and maintaining an updated log absence of such software standards has Go-HotSwap includes Jungo’s WinDriver, of the hot swap events according to the slowed down market growth. In addition, a driver development tool, as an integrated data received from the Hot Swap Engine. the lack of standards is forcing companies component. The driver development tool- It then transfers the relevant messages to to either give up using hot swap capabili- kit dramatically simplifies the very diffi- the subscribed drivers or applications. ties or develop or buy proprietary solu- cult task of developing a device driver from tions. Either way, the situation is a con- scratch. It provides a complete solution for Driver Development Toolkit siderable resource consumer. creating high performance drivers that han- This component, if you wish to use it, dle interrupts and I/O at optimal rates. enables quick and simple development of Another problem involves the hot swap your driver’s hardware access code. The standard itself, which is somewhat shady It is not obligatory to use the included driver development toolkit enables the in some of the software areas. The gray driver development toolkit in order to user to easily produce a user application areas in the specification have resulted in generate your driver’s source code. You aware of hot swap events. different implementations of some of the can create your driver from scratch and system board manufacturers’ hot swap simply add the Go-HotSwap API to your Configuration Manager/hs_activate mechanisms. As a result, a CompactPCI driver’s source code to make it hot swap This component is a user-mode appli- hardware vendor wanting to ship hot swap aware and utilize the Go-HotSwap ser- cation that activates and stops the Hot aware drivers with the hardware needs to vices. However, by using the included Swap Engine and enables the Hot Swap develop and provide a different driver for tools your development time is consider- Engine to respond to hot swap events. The each different system board requiring ably reduced. hs_activate is a console-mode application support. In order to provide a complete suitable for all supported operating sys- hot swap aware solution, the hardware Go-HotSwap also includes the Con- tems (Windows 2000/XP/Server 2003 and vendor needs to either use proprietary figuration Manager, a graphical tool, Solaris), while the Configuration Manager operating systems or implement the hot which enables the Go-HotSwap engine is a graphical interface for hs_activate, swap mechanism from scratch, and sepa- to respond to hot swap events. The user suitable for Windows (2000/XP/Server rately, for each system hardware type and can configure the Configuration Manager 2003) only. The user can configure the operating system needing support. in order to execute tasks upon hot swap Configuration Manager (or hs_activate) events such as running a batch file or start- in order to execute tasks upon hot swap A software infrastructure solution ing/stopping a service. In this way users events such as running a batch file or Jungo’s Go-HotSwap is a software infra- can achieve hot swap capabilities without starting/stopping a service, thus achieving structure that adds the necessary software modifying the driver’s source code. In hot swap capabilities without modifying modules required for CompactPCI hot addition, Go-HotSwap includes a sample the driver’s source code. swapping. The software also: in console-mode that has the same functionality as the Configuration Manager. The architecture diagram (Figure 1) illus- ■ Provides the tools and development The sample, hs_activate, also enables the trates two possible methods of working environment to develop hot swap Go-HotSwap engine to respond to hot with Go-HotSwap: aware drivers. swap events. ■ Complies with PICMG 2.1 R2.0 ■ The Configuration Manager is requirements for CompactPCI Jungo Go-HotSwap architecture responsible for activating the Hot General Use Full Hot Swap Go-HotSwap software is a combination Swap Engine. Software. of four components: ■ The Hot Swap Engine is then respon- ■ Supports Solaris and Windows 2000/ sible for detecting hot swap events XP/Server 2003, utilizing the native Hot Swap Engine and has the option of allocating the plug-and-play capabilities of these This component serves as a generic hot boards’ required resources. The operating systems. swap aware driver that resides in the ker- Hot Swap Engine notifies either the nel. The Hot Swap Engine communicates WinDriver application or the Con- The Go-HotSwap software infrastructure directly with the hardware, as well as with figuration Manager about hot swap consists of a low-level operating system the hardware specific driver (directly, or events. The Configuration Manager extension and a Hot Swap Engine respon- through the Configuration Manager). Its is also responsible for stopping the sible for: central task is to detect hot swap events, Hot Swap Engine. identify the hot swapped board’s required

36 / CompactPCI and AdvancedTCA Systems / November 2004 You can use WinDriver to produce the Conclusion Alexey Polonsky has been working user application, which includes auto- The hot swap feature has been a critical at Jungo’s Software Tools Division matic hot swap support. You have the requirement for application designs in the since August 2003 as Senior Software option to use the WinDriver wizard to telecom and datacom markets. The hot Developer. Alexey brings with him a generate your driver’s hardware access swap facility reduces or even eliminates strong background in software/hard- code. During this process, you are asked system downtime, which may be required ware connectivity, specializing in PCI to select your driver’s options, one of for replacement of boards and making and CompactPCI devices and solutions. which is plug-and-play notifications, changes in the system configuration. Before joining Jungo, Alexey served another is power management notifica- in an elite unit of the Israel Defense tions. Select these options in order to The combination of Go-Hot Swaps’s four Forces as a software engineer. He holds use WinDriver’s application with the Hot components enables hot swap capabilities a MSc in Electrical Engineering from Swap Engine. for CompactPCI hardware on all major the Technion, Haifa. operating systems and with any available An option is to use a different application system board. For further information, contact or service via the Configuration Manager, Alexey at: which notifies, starts, or stops the driv- Go-HotSwap enables users to immedi- ers or applications according to the way ately add hot swap capabilities to their Jungo Software Technologies the user configured the Configuration existing PCI drivers or to easily develop 3031 Tisch Way Manager. hot swap aware drivers from scratch. Suite 1007 The resulting code will compile and run San Jose, CA 95128 Using the software on all supported operating systems. Go- Tel: 877-514-0537 There are two basic methods in which HotSwap also transparently supports any Fax: 877-514-0538 Go-HotSwap can be used. CompactPCI hardware, such as basic or E-mail: [email protected] full hot swap. It is also silicon friendly Website: www.jungo.com Using the WinDriver/ and ready. Go-HotSwap API Use Jungo’s driver development tools to develop a driver from scratch, that will automatically support hot swap events.

Using the Configuration Manager Configure the system to execute pre- defined tasks upon hot swap events according to a user-defined set of criteria. This method enables hot swap capabilities with legacy drivers without the need to modify the driver’s source code, provided that the driver complies to the PCI dynamic loading/unloading standard.

Figure 1 RSC# 37 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 37 TECHNOLOGY F E A T U R E :

MILITARY Israel Defense Forces deploy AdvancedTCA core routing switch

By Zvi Marom

he Israel Defense Forces Deployments began subsequent to a Training and educational curves both (IDF) recently became the proof-of-concept trial that, among other figure very prominently in the T6Pro’s initial organization to deploy things, demonstrated switching vendor low cost of ownership. Currently, network Tan operational and commercially avail- interoperability. In this trial several T6Pro operators using proprietary platforms for able AdvancedTCA based Layer 3 platforms served as the main aggrega- the various network and infrastructure routing switch platform. In this article tion layer for a LAN network, which is layers (such as IP routers, video servers, Zvi addresses the important role that operational today and includes numerous blade servers, ATM switches, and many AdvancedTCA architecture and the aggregation and access routing switches others) must utilize multiple training and AdvancedTCA based T6Pro switch is from Telco Systems and other switch educational resources and generally are playing in the advanced voice, video, vendors. required to pay excessive prices for this and data networks that are now being proprietary training. deployed. In the IDF deployment, the T6Pro is connected to the Core IP-MPLS layer (core Proprietary platforms are also often An AdvancedTCA system core routing router based) and functions as the gate- accompanied by a small army of techni- switch called the T6Pro (Figure 1), was way between the WAN and LAN layers. cians. Moreover, the different layers and developed by Telco Systems for carrier- The T6Pro and the other Telco Systems applications typically demand several sup- class, multigigabit routing and switching platforms, which include various mem- porting teams. This results in the need for aggregation applications. AdvancedTCA’s bers from its T5C family of compact extremely high professional qualifications objective is to provide an open, multiven- routing switches, run L2-VPN and for technicians, who must be aware of the dor standardized architecture for car- L3-VPN as well as advanced quality very latest on specific technologies, as well rier-grade applications. AdvancedTCA of service and security features. T6Pro as knowing the details of products from defines blade and shelf types optimized blades support MPLS features in hard- different vendors. Even skilled technicians for telecommunications and guarantees ware, providing this value-added capabil- require a considerable investment by the vendor interoperability. ity as an intermediate solution targeted at network operator to keep them at the top LAN-WAN and MPLS networks. of the learning curve. AdvancedTCA standardization, as demonstrated in the T6Pro, Telco Systems’ T6Pro was chosen over simplifies training needs and broadens the Cisco’s equivalent platform primar- area that a single technician can address, ily for cost/performance reasons that which results in much higher levels of effi- can be attributed, at least in part, to the ciency for every technician. AdvancedTCA concept with its standardized platform that provides pricing advan- The T6Pro’s no single-point-of-failure tages that are hard to ignore. AdvancedTCA architecture, port density advantage, and transceiver type possi- The T6Pro’s cost of ownership (including bilities were also targeted specifically initial equipment cost, installation, rou- by the IDF because it provided for maxi- tine operation, retrofit, and decommis- mum flexibility and field upgradeability. sioning costs) has been found to lower the Telecom companies and enterprises that costs an estimated 60-70 percent over a are unwilling to compromise and insist on period of five years compared to today’s carrier-class features require such heavy- available architectures. duty capabilities as: Figure 1 Standardization lowers costs by freeing ■ Very high reliability the network operator from dependence on ■ Scalability A number of months ago, Telco Systems vendor-proprietary platforms. In addition, ■ Standardization to ensure interoper- initiated deployments of the T6Pro once the platform and its building blocks ability and to guarantee independence AdvancedTCA core routing switch along are standardized and available from sev- from single source vendors with some of its other IP product line eral equipment and solution vendors, this platforms as part of the IDF’s nationwide results in lower inventories and decreased The T6Pro’s enhanced security, quality of fixed-line telecommunications project. storage costs. service, end-to-end provisioning manage-

38 / CompactPCI and AdvancedTCA Systems / November 2004 ment, and sub 50 ms switchover times are T6Pro Classic and T6Pro Compact, which consultant to several industrial and additional features essential for superior use the same card types. The Classic offers academic institutes. He holds degrees in carrier-class operation. The T6Pro sup- a 14-slot architecture, and the Compact Engineering and Medicine and gradu- ports metro access applications and pro- has a 5-slot architecture. For appli- ated with excellence from the Naval vides multigigabit aggregation services. Its cations with up to 96 GigE ports, the Academy and served in combat com- carrier-class reliability and availability fig- T6Pro Compact provides a cost-effective mand posts. ures guarantee 99.999 percent uptime, and solution, while for more port intensive the system’s capacity and scalability fits applications the T6Pro Classic is prefer- For further information, contact Zvi or current and future carrier requirements. In able. Bo Sullivan at: addition, the T6Pro uses a redundant CPU and Matrix (CPM) architecture, provid- Zvi Marom is CEO of Telco Systems Telco Systems ing an industry record sub 50 ms failover and the founder and CEO of its parent 2 Hampshire Street, Suite 3A switching time. All cards are hot swap- company, BATM. He started BATM in Foxboro, MA 02035-2897 pable and are powered by individual on 1992 and acquired Telco Systems in Tel: 800-221-2849, ext. 5286 board -48 VDC power supply modules, 2000. Prior to establishing BATM, Zvi E-mail: [email protected] or making for a power distribution system was the head of the electrical faculty [email protected] with no single point of failure. of Israeli Open University and senior Website: www.telco.com

The T6Pro also allows services conver- gence with IP/LAN and wireless, server blades, transport, ATM, VoIP, and others by using AdvancedTCA-compliant modules. For example, an AdvancedTCA- based PC blade could be plugged in and used as a management station.

The T6Pro’s AdvancedTCA architecture allows for near immediate services implementation, using a COTS platform. Its low entry and maintenance costs represent the lowest cost-of-ownership available today. Another plus is its full conformance with carrier requirements for NEBS, DC feed- ing, and other carrier-class features.

With the T6Pro platform, Telco Systems offers an extensive range of line cards including 10-Gigabit Ethernet, Gigabit Ethernet, and Fast Ethernet. Its switching fabric provides wire-speed nonblocking connectivity of more than 280 Gigabit Ethernet ports in a single 12RU chassis, making it the densest Gigabit platform on the market. Telco Systems enclosure per- mits the installation of more than 1,000 GigE optical ports in a single 19-inch rack. While metro IP platforms require twice as much rack space, the T6Pro accommodates the same number of ports in half the space.

Furthermore, the T6Pro offers a unique AdvancedTCA-compliant 10 Gbps Net- work Processor (NP) module. While most application implementations require long lead times, T6Pro’s NP module provides a flexible solution today. IPv6 dual-stack support, firewalls, or any other software application can also be integrated quickly as more and more vendors are developing applications for AdvancedTCA based platforms.

For size and capacity optimization, Telco Systems offers two platform types, the RSC# 39 @www.compactpci-systems.com/rsc

CompactPCI and AdvancedTCA Systems / November 2004 / 39 TECHNOLOGY F E A T U R E :

HIGH PERFORMANCE Building high performance edge/core switches using an AdvancedTCA platform – implementation issues

By Reuven Segev

uilding high performance Core router issues in an based backplane (Figure 2). Selecting edge/core switches presents AdvancedTCA platform which scheme to use hinges on an eco- interesting challenges to design- A different use for an AdvancedTCA nomic engineering decision driven by the Bers using AdvancedTCA platforms. In platform is implementing core switch/ cost of switch cards and the potential loss this article Reuven analyzes a number routers. Gilbert and Rasovsky classified of revenue generating blades. It is also of issues related to the selection of these as “Routing (Transit) Applications.” driven, as indicated later in this article, by a switch fabric able to operate in an In order to be considered as carrier-class the wire speed requirement. AdvancedTCA environment and offers systems, core router/switch applications a solution to this problem. are required to deliver wire speed perfor- The wire speed requirement is particularly mance at 10 Gbps user bandwidth and challenging in an AdvancedTCA standard The emergence of AdvancedTCA as the support high resiliency. Furthermore, platform. The AdvancedTCA documents platform of choice for new telecommuni- the current CapEx environment strongly specify a 10 Gbps connection between cations and data communications systems emphasizes achieving these goals with each node (line card) and the hub (switch is a fact, and at this time requires no addi- little or no impact on systems’ cost. card). This 10 Gbps connection is imple- tional supporting arguments. The positive mented by means of 8 unidirectional links market dynamics of the AdvancedTCA The high resiliency requirement can be operating at the rate of 2.5 Gbps each. In ecosystem do not require further argu- met by support of 1:1 or 2:1 protection practice, backplane vendors qualify their ment, and as more vendors join the schemes that guarantee circumventing a backplanes to 3.125 Gbps rate. This sup- AdvancedTCA community, the question single point of failure. 1:1 protection is ports a raw data rate of 12.5 Gbps bidi- of whether AdvancedTCA will take off implemented using Dual Star architecture rectional data. At first glance this seems has been replaced by how soon will it (see Figure 1). In addition, a 2:1 protec- to satisfy the 10 Gbps wire speed require- happen and which applications will ben- tion scheme mandates the use of a mesh ment. That is not the case, however, when efit most. This article discusses the use of AdvancedTCA in edge/core router/switch applications.

AdvancedTCA based platforms have made their debut in emerging wireless applications in edge and service areas of the network, such as Radio Network Controllers (RNCs), Serving GPRS Support Nodes (SGSNs), and Home Location Registers (HLRs). Gil- bert and Rasovsky classify these applica- Figure 1 tions as “Application Processing (Trans- formation) Applications.”[1] They are characterized by multihop connections whereby an incoming packet can be processed by multiple nodes in the system, passing a number of hops before egressing to the network. This type of platform is also characterized by a relatively modest backplane bandwidth capacity. Blade server applications are similar in nature to Service Nodes, and several companies are evaluating opportunities in this segment. Figure 2

40 / CompactPCI and AdvancedTCA Systems / November 2004 one takes into account all Physical Layer wire speed performance for 10 Gbps user posals have been submitted to IEEE and and Link Layer overheads. traffic. other technical forums, and a number of established as well as start-up vendors are The first factor to reduce the net data rate The requirement for eight lanes between aggressively promoting it. Nevertheless, is the Physical Layer encoding. In order each node and the active hub violates the the technical problem is complex, and the to guarantee the integrity of the data and AdvancedTCA specification, which sup- standardization process is probably a year the synchronization among the four lanes, ports only four such lanes. Hence, an or more from completion. data is encoded using an 8B/10B code. AdvancedTCA based platform appears The result is a 20 percent Physical Layer to be unsuitable to support core routers/ Meanwhile, a number of measures can be overhead that has the result of reduc- switches required to support line cards taken in order to improve the efficiency ing the net data rate from 12.5 Gbps to with 10 Gbps user traffic. of link bandwidth utilization and enable 10 Gbps. AdvancedTCA to be used as a platform of Solving the speed deficit choice for core routers/switches. The next factor affecting the total avail- Future revisions of the AdvancedTCA able bandwidth is the Link Layer over- specification address this problem by The first measure reduces the cell tax or head. While no protocol has been stan- migrating to higher speed SERDES, op- Link Layer overhead. The way to do that dardized for the line card to switch card erating at 6.25 Gbps or higher. Using is to use variable-size cells. Using vari- link, all existing solutions are cell based. 6.25 Gbps, the node-hub link supports able-size cells can minimize both fixed The use of cell based Link Layer proto- 25 Gbps on four lanes, thus providing the overhead and fragmentation overhead. A col for the transport of packets requires necessary speedup. packet is segmented into large-size cells – the encapsulation of packets in cells. minimizing the fixed overhead – with the This Segmentation and Reassembly Work is in progress on the standardization last cell being selected such that it has the (SAR) process introduces two types of high-speed SERDES, but the process is minimum size possible – minimizing the of overhead: a static overhead consist- still in its infancy. Different technical pro- fragmentation overhead (Figure 4). Such a ing of the additional cell overhead and a fragmentation overhead generated in the case where the transported packet is not an integer number of cells (Figure 3). The trade-off is using small cells (small fragmentation overhead and large static overhead) or large cells (small static overhead and large fragmentation overhead). In both extreme cases, the use of cells creates a need for excess bandwidth and thus speedup. The speedup required by the Link Layer overhead is on the order of x1.4. Thus, 10 Gbps user traffic requires a link rate of 17 Gbps, when taking into account also the 8B/10B encoding.

An additional factor impacting the wire speed capability of the router relates to the switch fabric’s architecture. Switch Figure 3 fabrics are based on either cross-bar or shared memory architectures. In a cross- bar based switch fabric, a request-grant protocol is used between the ingress VoQs and the scheduler (centralized or distributed scheduler architectures are used by different vendors). The request-grant protocol generates additional traffic, resulting in a need for additional algorithmic speedup of x1.2.

To summarize: A shared-memory based system requires a speedup ratio of about x1.4, resulting in an overall link rate of 17 Gbps. A cross-bar based fabric requires a speedup of about x1.7 resulting in a link rate of 21 Gbps.

To provide the necessary speedup, commercial fabrics use eight lanes with a total capacity of 25 Gbps in order to guarantee Figure 4

CompactPCI and AdvancedTCA Systems / November 2004 / 41 TECHNOLOGY F E A T U R E : H I G H PERFORMANCE system, built by TeraChip, has been tested is supported at 10 Gbps by utilizing both tem can be operated in a 2+1 protection in real life applications for a wire speed switch cards (hubs). In the event of sched- mode. In this mode of operation, the sys- performance of 8 Gbps on only four lanes uled maintenance or in case of a hub fail- tem includes three switch cards and the operating at 3.125 Gbps per lane. ure, the system will continue to operate traffic emanating from a line card is split with a single hub, retaining the capability among all three cards. In the event of a An additional measure to be taken is to support 8 Gbps wire speed. switch card failure, the traffic is automati- supporting a 1+1 (load sharing) protec- cally split between the two operational tion mechanism. In this case, the system Certain high performance applications re- switch cards, retaining the necessary band- operates in a Dual Star configuration with quire support of full wire speed at 10 Gbps width required to support full 10 Gbps both switch cards operational and the traf- at all times, including in the presence of wire speed capability. fic split between the two. Full wire speed failures. In support of such cases, the sys- In conclusion, we have demonstrated that a solution exists today for an efficient implementation of a core router/switch using current versions of AdvancedTCA based chassis.

[1] J. Gilbert and K. Rasovsky, “Choosing the Right ATCA Fabric,” CommsDesign, August 3, 2004

Reuven Segev is Director of Marketing at TeraChip, Inc. Reuven has a long track record of product management, marketing, and business development in different areas of telecommunications with Trellis Photonics, OrAccess, Ikanos Communications, and Diamond Lane Communications. Prior to that, Reuven had a long and distinguished career in Engineering with ECI Telecom. He holds a BSEE degree from the Technion in Israel.

For further information, contact Reuven at:

TeraChip, Inc. 2479 East Bayshore Rd Suite 700 Palo Alto, California 94303 Tel: 650-320-8148 Fax: 650-320-8149 E-mail: [email protected] Website: www.tera-chip.com

COMING

DECEMBER 2 0 0 5

BUYER’S

GUIDE

RSC# 42 @www.compactpci-systems.com/rsc

42 / CompactPCI and AdvancedTCA Systems / November 2004 CompactPCI andRSC# AdvancedTCA 43 @www.compactpci-systems.com/rsc Systems / November 2004 / 43 I/O PRODUCTS I/O PRODUCT GUIDE

Zendex ZXE-UFE/104P

Company Name Website Analog Digital Industrial Multifunction

Acquisition Technology • • www.acq.nl Acromag • • • www.acromag.com ACT/Technico • • www.acttechnico.com Actis • • www.actis-computer.com RSC 15021 Adas • • www.adas.fr Tews Technology LLC Adcom • www.adcomtec.com TPMC810 ADDI-DATA • • www.addi-data.com Adlink Technology • • • www.adlinktech.com Advantech • www.advantech.com Agilent • www.agilent.com Aitech • www.rugged.com RSC 18051 Alacron • www.alacron.com Alphi Technology • • • www.alphitech.com Meilhaus Electronic GmbH ME-1600 Analogic • • www.analogic.com Annapolis Micro Systems • www.annapmicro.com Axiom Technology • • www.axiomtek.com BiRa Systems • www.bira.com BittWare • • www.bittware.com BVM • • www.bvmltd.co.uk BWI • • • • www.bwi.com RSC 15594 C&H Technologies • • • www.chtech.com CES • www.ces.ch Measurement Computing PMD-1024HLS Comtel Elect • www.comtel-online.de Concurrent Technologies • • www.gocct.com Condor Engineering • www.condoreng.com Curtiss-Wright Controls • • • www.cwcembedded.com Cyclone Microsystems • www.cyclone.com Data Design Corporation • www.datadesigncorp.net Datel • • • www.datel.com Diversified Technology • www.dtims.com DNA Enterprises • www.dna-cs.com

RSC 17778 DRS • www.drs.com

44 / CompactPCI and AdvancedTCA Systems / November 2004 I/O PRODUCT GUIDE

Company Name Website Analog Digital Industrial Multifunction

DSPCon • www.dspcon.com Dynamic Engineering • • www.dyneng.com Echotek • www.echotek.com EKF-Electronik • www.ekf.de ELMA Electronic • www.elma.com esd • • www.esd-electronics.com FuturePlus Systems • www.futureplus.com Galil Motion Control • www.galilmc.com GE Fanuc Automation • • • • www.gefanuc.com/embedded General Standards • • • • www.generalstandards.com Geotest • • www.geotestinc.com ImageStream • www.imagestream.com Inova • • www.inova-computers.de Interactive Circuits & Sys. • • www.ics-ltd.com Interface Amita • www.interface-co.com IOtech • www.iotech.com Joerger Enterprises • www.joergerinc.com Kontron • • • www.kontron.com Maxwell Technologies • www.maxwell.com Measurement Computing • • • www.measurementcomputing.com MEI • www.motioneng.com Meilhaus Electronic • • www.meilhaus.com MEN Micro • • • • www.menmicro.com Mercury Computer Systems • www.mc.com Merlin Electronics • www.merlinelectronics.com N.A.T. • www.nateurope.com National • www.national.com National Instruments • • www.ni.com Netbricks • www.netbricks.net NEXCOM International • www.nexcom.com NMS Communications • www.nmscommunications.com

Continued on page 47

RSC# 45 @www.compactpci-systems.com/rsc 46 / CompactPCI and AdvancedTCA SystemsRSC# / November 46 @www.compactpci-systems.com/rsc 2004 I/O PRODUCT GUIDE

Company Name Website Analog Digital Industrial Multifunction

Pentek • • • www.pentek.com Precision Communications, Inc. • • www.precisioncomm.com Radstone Technology • • www.radstone.co.uk Rittal • www.rittal.com RT Logic! • www.rtlogic.com Rymic • • www.rymic.com SBS Technologies • • www.sbs.com Schroff US • www.schroff.us Sensoray • • www.sensoray.com SMA • www.SMAcomputers.com SMI • www.sonoranmicrosystems.com Soltec • • www.solteccorp.com Spectrum Sig. Proc. • • • www.spectrumsignal.com Strategic Test • • • www.strategic-test.com Sundance DSP • • www.sundance.com Synergy Microsystems • www.synergymicro.com Technobox • • www.technobox.com Technology Dynamics • www.technologydynamicsinc.com TEK Microsystems • • www.tekmicro.com Tenta Technology • • www.tenta.com Tews Technologies • • • • www.tews.com Thales Computers • • www.thalescomputers.com Transtech DSP • • www.transtech-dsp.com Traquair Data Systems • www.traquair.com United Electronic Industries • www.ueidaq.com Varisys Ltd • www.varisys.co.uk Vigilant Technologies • www.vigilanttech.com VMETRO • www.vmetro.com Woodhead • www.woodhead.com Zendex • www.zendex.com Zephyr Engineering • www.zpci.com

RSC# 47 @www.compactpci-systems.com/rsc RSC# 48 @www.compactpci-systems.com/rsc 48 / CompactPCI and AdvancedTCA Systems / November 2004 NEW PRODUCTS By Chad Lumsden E-mail: [email protected] CompactPCI

P1386.1 • PCI 2.1 compliant interface • 149mm x CHIPS & CORES: IP Tews Technologies 74mm • SERCOS interface based on SERCON410 GDA Technologies Website: www.tews.com chip • Single chip controller for SERCOS interface • Website: www.gdatech.com Model: TPMC862 RSC No: 18059 Dual Port RAM 1024 words x 16 bit • Transmission Model: GRIO RSC No: 17752 A standard single-width 32 bit PMC with four rate up to 4 Mbaud with internal clock recovery • RapidIO controller IP cores • Fully compliant high speed serial data communication channels Internal repeater for ring connections • Full duplex with the Serial and Parallel RapidIO specifica- conforming to IEEE P1386.1 • PCI 2.1 compliant operation • Physical interface optical fiber ring or tions • Derived from the RapidIO technology- master/slave interface • 147mm x 74mm • Four RS-485 • Two encoder inputs RS-422 or TTL for based controller design licensed from Freescale high speed synchronous/asynchronous serial hand wheel functionality Semiconductor • Optimally supports Standard interfaces based on Infineon PEF20534 • Support Cell, FPGA, and Gate Array • Independent of appli- of RxD, TxD, RxC, TxC, RTS, CTS, CD, and GND Model: TPMC813 RSC No: 18053 cation logic, PHY designs, and target technolo- on HD68 front connector, parallel to rear connector A standard single-width 32 bit PMC module pro- gies • IP package consists of Verilog RTL code, a P14; DTR3 and DSR3 only at front I/O • Physical viding a complete LON interface using the Neuron verification environment, test cases, and detailed interface (individually programmable per channel): chip 3150 • Conforms to IEEE P1386.1 • PCI 2.1 design documentations • Design services avail- EIA-232, EIA-422, EIA-449, EIA-530, EIA-530A, compliant interface • 149mm x 74mm • LON inter- able for customization and integration of the IP V.35,V.36 and X.21 • 10 Mbps (synchronous), face based on Neuron chip 3150 • 4 KB dual port cores into each customer’s ICs and FPGAs 2 Mbps (asynchronous) maximum data rate, internal RAM • 32 KB EPROM for protocol firmware and or external provided clock • Up to 115.2 kbps EIA-32 application • Physical interface RS-485 trans- • -40°C to +85°C operating temperature range ceiver or transformer coupled (1.25 Mbit or 78 DATACOM: ETHERNET Kbit) and FTT-10A transceiver • Complete network Tews Technologies communication port of the neuron chip available ENCLOSURE + CARD RACK + at front panel • MIP/DPR license included Website: www.tews.com POWER SUPPLY Model: TPMC881 RSC No: 18069 Model: TPMC821 RSC No: 18054 A standard single-width 32 bit PMC module One Stop Systems A standard single-width 32 bit PMC module offer- providing a single channel Ethernet 10Base-T/ Website: www.onestopsystems.com ing a complete PMC INTERBUS Master Generation 100Base-Tx interface conforming to IEEE P1386.1 Model: 4U-19S-2-350 RSC No: 17773 4 (G4) interface • Conforms to IEEE P1386.1 • • PCI 2.1 compliant interface • 147mm x 74mm A passive backplane-based rackmount enclosure PCI 2.1 compliant interface • 149mm x 74mm • • IEEE 802.3 compliant LAN interface • Support for applications with high shock and vibration Local MCU Motorola MC68332 20 MHz • Interbus of front panel status LEDs • Half or full-duplex requirements • Supports a wide range of passive Controller IPMS3 Interbus Master Protocol Chip operation • 3 KB transmit and receive FIFOs • backplanes, including segmented backplanes, • One MB SRAM • One MB Flash plus 256 kbits Controller supports DMA cycles as a bus master supporting multiple systems in a single enclosure x 16 • 4K x 8 DPM • Interbus interface (front I/O) • Temperature range: 0°C to +85°C (TPMC881-10/ • Five high-output fans • 3-1/2" and 5-1/4" modular DB9 female connector • RS-232 interface (front I/ TPMC881-11); -40°C to +85°C (TPMC881-10-ET/ drive bays • Dual hot swappable 350-watt power O) shielded 8P. Mod.-Jack • LED diagnostic (front TPMC881-11-ET) supplies • Supports 19-slot PICMG-compatible I/O) • Enhanced diagnostic port at the P14 mez- passive backplanes • Removable backplane tray • zanine connector (back I/O), RESET push-button DATACOM: SERIAL CONTROLLER Reinforced corners • Power supply hold-downs • (front I/O) • G4 Phoenix Contact firmware (4.66) Flush-mounted LEDs • Completely secured cables Tews Technologies Website: www.tews.com I/O: DIGITAL Model: TPMC362 RSC No: 18057 FIELDBUS: CAN Measurement Computing A conduction cooled PMC four-channel high Tews Technologies Website: www.measurementcomputing.com speed Synch/Asynch Serial Interface • Standard Website: www.tews.com Model: PMD-1024HLS RSC No: 17778 single-width 32 bit PMC module conforming to Model: TPMC316 RSC No: 17768 A USB–based analog and digital I/O interface IEEE P1386.1 • PCI 2.1 compliant master/slave Conduction cooled single-width 32 bit PMC mod- • Fully compatible with USB 1.1 as well as USB interface • 147mm x 74mm • Four high speed ule conforming to IEEE P1386.1, no front panel • 2.0 • Provides 24-bits of high-current digital I/O • synchronous/asynchronous serial interfaces PCI 2.1 compliant interface • Board size: 144mm Up to 15mA of source and 64mA of sink current • based on Infineon PEF20534 • Support of RxD, x 74mm • Two CAN bus interfaces based on Intel High currents eliminate the need for external buf- TxD, RxC, TxC, RTS, CTS, CD, and GND on rear 82527 chip • Support CAN specification 2.0 part fers and allow the PMD-1024HLS to drive many connector P14 • Physical interface (individually A and B (standard and extended data frames) • external devices directly • Digital I/O is based on a programmable per channel): EIA-232, EIA-422, Programmable global mask • 15 message objects discrete implementation of the industry standard EIA-449, EIA-530, EIA-530A, V.35, V.36, and of 8 byte data length • Powerful error handling • 82C55’s mode • 24 DIO bits are divided into two X.21 • Maximum data rate: 10 Mbps (synchro- Programmable transfer rates • Physical interface 8-bit ports and two 4-bit ports, each of which may nous), 2 Mbps (asynchronous), internal or exter- CAN High Speed (according to ISO 11 898) on be independently programmed as input or output nal provided clock EIA-232: up to 115.2 kbps • TPMC316-10 or modified RS-485 on TPMC316- • 3.25" (W) by 3.125" (L) by 1" (H) • Magnetic and Temperature range: -40°C to +85°C 11 per channel • Physical interface optically Velcro mounting options • .225" attachment hole • ▲ isolated from CAN controller by onboard DC/DC Attachment hole can also be used to attach a cable converter and opto coupler per channel • Transfer lock or padlock in applications where unit security rate 1 Mbps maximum • Operating temperature must be protected • Windows 98SE/ME/2000/XP range -40°C to +85°C operating systems

FIELDBUS: OTHER I/O: INDUSTRIAL Tews Technologies Tews Technologies Website: www.tews.com Website: www.tews.com Model: TPMC812 RSC No: 18052 Model: TPMC310 RSC No: 18056 A standard single-width 32 bit PMC module with a A conduction cooled single-width 32 bit PMC RSC 18057 complete SERCOS bus interface conforming to IEEE module providing two channels of high speed

CompactPCI and AdvancedTCA Systems / November 2004 / 49 CAN bus interface • Conforms to IEEE P1386.1 count frequency: 1.2 MHz • Digital signal filter Serial Port • 2 USB Ports • 3 on-board FC Drives • PCI 2.1 compliant interface • 144mm x 74mm for suppression of fast noise pulses • Any origin for up to 210 GB of storage • ACPI BIOS • Two channel high-speed CAN interface • Two within driving range selectable • Auto-reference Philips SJA1000 CAN controller • Supports procedure • 16 bit D/A converter per axis with CAN specification 2.0B • I/O interface based on ±10V output signal to control servo amplifier • 12 SYSTEM MANAGEMENT TJA1050 transceiver, channels isolated from sys- bit A/D converter per axis with adjustable input Augmentix tem logic and from each other • On board termina- voltage range, e.g. for connection of a joystick Website: www.augmentix.com tion option for each CAN bus channel • Operating • Four inputs per axis for reference switch, limit Model: A+ SAMP RSC No: 17761 temperature range –40°C to +85°C switches, emergency stop • Three outputs per PCI management and alarm card • Can be used axis for control of axis enable and current limit- in a wide-number of commercial servers to pro- Tews Technologies ing • TIP102-TM model has: 50 pin flat cable con- vide in-band and out-of-band platform manage- Website: www.tews.com nection to TIP102; signal distribution by three DB ment for High Availability • Allows for indepen- Model: TPMC810 RSC No: 18051 connectors per axis in a 6U / 1-slot front panel; dent monitoring and control of servers using the A standard single-width 32 bit PMC with two inde- encoder interface directly to RS-422 or TTL level, SA Forum HPI interface • Provides standardized pendent CAN bus channels, isolated from system and makes possible signal conditioning of all lights-out management across a mix of servers logic and from each other • Standard single-width signals from multiple manufacturers • Server Class plat- 32 bit PMC module conforming to IEEE P1386.1 • form management • 266 MHz PowerQUICC II pro- PCI 2.1 compliant interface • Board size: 147mm x PCMCIA/CARDBUS cessor • 128 MB of main memory • Up to 1 GB 74mm • Two CAN bus interfaces based on Philips of removable storage • Two MB of boot storage SJA1000 • I/O access: DB9 male front connec- Tews Technologies and diagnostic firmware • Dual 10/100 external tors with pinout following CiA DS-102, parallel Website: www.tews.com Ethernet • Operating system agnostic Ethernet to rear connector P14 • Physical interface: CAN Model: TPMC872 RSC No: 18063 based interface to host • USB 1.1 I/O for modem High Speed (according to ISO 11898-2) • Physical A standard single-width 32 bit PMC module pro- and other USB supported devices • Dual exter- interface optically isolated from CAN controller by viding one or two sockets for 16 bit PC Card or 32 nal power for continued operation even if server on board DC/DC converters and optocouplers for bit CardBus Cards using a dual socket PC Card power fails • Separate instrumentation processor each channel • Transfer rate 1 Mbps maximum • Conforms to IEEE P1386.1 • PCI 2.2 compliant • Temperature monitoring of up to eight points (bus length up to 30m) • Temperature range: -40°C interface • 149mm x 74mm • Up to two PC Card/ – four thermocouples • Fan speed monitoring and to +85°C CardBus Sockets with hot insertion and removal, control for up to 16 server fans • Server power supports memory and I/O cards of type I, II, and monitoring • Server power and reset control • I2C III (second socket only) • -40°C to +85°C operat- connection to System Management Bus (SMBus) I/O: MULTIFUNCTION ing temperature or IPMI native implementations Tews Technologies Website: www.tews.com PROCESSOR: POWERPC ▲ Model: TPMC851 RSC No: 18058 A standard single-width 32 bit PMC module con- MEN Micro forming to IEEE P1386.1 • Target chip: PCI9030 Website: www.menmicro.com (from PLX Technology) • PCI 2.2 compliant inter- Model: EM04N RSC No: 17979 face • PCI I/O signaling voltage 5V and 3.3V • 32 A complete embedded single board computer for channels single-ended or 16 channel differential use on any carrier board • MPC8245 PowerPC 16 bit multiplexed analog input • Programmable (Kahlua II) microprocessor at 266 MHz to 400 gain amplifier (gain 1, 2, 4, or 8) • Full-scale input MHz • SODIMM socket for RAM • CompactFlash range: ±10V (at gain 1) • Minimum 1.25 µs, maxi- slot for program storage • Two RS-232 serial mum 17.25 µs conversion time depending on interfaces and two Fast Ethernet interfaces on the RSC 17761 mode • 70Vpp overvoltage protection • 8 chan- front panel • Onboard Cyclone FPGA with 250,000 nels 16 bit analog output • ±10V output voltage • gates Conversion time: 10 µs • Up to 10,000pF capaci- tive load • Sequencer • 16 digital TTL I/O lines with TEST SYSTEMS pull up resistors • Individually programmable as PROCESSOR: XEON Acme Portable Machines input or output • Programmable debounce time Diversified Technology Website: www.acmeportable.com (100ns – 6.55ms) • Interrupt capable • 32mA Website: www.dtims.com Model: GCP382 RSC No: 17869 source/sink • 4.7 kohm pull-up resistor • ESD and Model: ATC-4130 RSC No: 17967 High impact resistant plastic, and high strength overvoltage protected • 32 bit multipurpose coun- Single LV 2GHz Intel Xeon Processor • Intel E7501 aluminum plate chassis • Complies with PXI ter • 32 bit preload register • 32 bit compare reg- Chipset • 4 GB Memory • Two 64bit/66Mhz PMC specifications • Support 3U PXI 8 slot backplane ister • Various count and control modes • Count with PCI-X support • GB Ethernet Base Interface • • Power Supply: Full range ATX 350W • Four 8cm frequency: External clock up to 10 MHz; internal GB Ethernet Fabric Interface (ATCA 3.1 Option 1) • Fans, air flow up to 120 CFM • 16.5" (w) x 10.8" (l) clock 5, 10, 20, or 40 MHz • Factory calibrated, Optional Fibre Channel Interface (ATCA 3.1 Option 4) x 8.3" (d) 419mm (w) x 275mm (l) x 210mm (d) calibration data stored in EEPROM • Operating • Front Panel 10/100 Ethernet Port • Front Panel • Internal flat panel 14.1" or 15" XGA TFT display temperature -40°C to +85°C ▲ • Touch Screen MOTION CONTROL Tews Technologies Website: www.tews.com For further information, Model: TIP102/102-TM RSC No: 18090 An IndustryPack compatible motion controller enter the product’s RSC# at using incremental encoder • Interface according to IndustryPack specification • Identification- www.compactpci-systems.com/rsc PROM supports auto-configuration • Single size IndustryPack • 24 bit counter per axis (± 8.300.000 increments) • Programmable single, double, or RSC 17967 quadruple analysis of count signals • Maximum

50 / CompactPCI and AdvancedTCA Systems / November 2004 RSC# 51 @www.compactpci-systems.com/rsc RSC# 52 @www.compactpci-systems.com/rsc