AdvancedMC Insider February 2006

As demand for higher bandwidth increases, highly specialized equipment that can economically and flexibly handle wire speed data and voice traffic is starting to appear. Most of this telecommunication-specific equipment is based on network processor technology. Before the advent of dedicated network processors, network traffic was handled by equipment based on general-purpose or ASIC processors, which carry unneeded hardware and software overhead, and ASICs, which are specific to the task of network processing, but are not upgradeable, take a long time to design, can't be reprogrammed and are economical only in very high volumes.

Dedicated network processors, on the other hand, are relatively inexpensive programmable processors that are optimized for processing network packets efficiently and are designed with performance features for the higher speed networks that are increasingly based on Gigabit Ethernet. Some of the advantages of network processors include programmability, performance, management, and routing.

Network processors can provide network traffic analysis functions like examination of data packets to determine how the packets will be processed and forwarded. For example, a network processor may be used to examine data packets using a programmed algorithm that keeps track of system bandwidth and regulates contracted bandwidth speeds.

The advantages of using a network processor are immediately apparent when facilitating the migration of network services from legacy networks to IP-based convergent gateway and ATM to IP interworking applications like convergent 3G wireless, Voice over Internet Protocol (VoIP), Media Gateways, DSLAMs and Switch/Routers.

Considering that the open standard AdvancedTCA® and MicroTCA^TM platforms are targeted at telecom applications, it was only a matter of time before network processors began to appear on the AdvancedMC modules that give those platforms their individual personalities. Using a network processor to do the heavy lifting rather than a more expensive general processor-based computer makes high-performance packet processing more affordable—a major factor when dealing with large application distribution.

Take the example of a MicroTCA system using a network processor-based AdvancedMC module acting as a gateway between an ATM-based network and a switched fabric Gigabit Ethernet backplane. This one card enables system designers to employ a single module to build high-performance communications systems using a distributed processing architecture. The AdvancedMC module can directly terminate and switch traffic on the card rather than sending traffic to an off-board processor, thus reducing design complexity, saving cost and boosting performance in the bargain.

The performance advantage of a network processor is why SBS Technologies® uses one in its Telum 1204-O3 AdvancedMC module. The Telum® 1204-O3 is the industry's first intelligent high performance, cost effective multi-service OC-3 interface module in AdvancedMC format offering high-end ATM and IP services based on a state-of-the-art Wintegra® WinPath® network processor. It provides termination, switching and interworking capabilities from any port to any port on a single card. For more information on the Telum 1204-O3 click here.

Rubin Dhillon, V.P. Communications & Enterprise The AMC.0 specification defines a serial fabric interconnect between the AdvancedMC card and its carrier, but the specification does not mandate what the fabric should be. Subsidiary specifications AMC .1 through AMC .4 define several possibilities, and there has been a certain amount of discussion as to which fabric should be employed.

As is usual with these things, there is no “best” or “one size fits all.” Before discussing the relative merits of each option, it is important to remember that the AdvancedMC^TM card can be used either on a carrier for AdvancedTCA or IBM BladeCenter®T systems or as a kind of “mini-blade” in the emerging MicroTCA^TM systems. This dual usage scenario complicates the situation, because the fabric that might make the most sense for the MicroTCA backplane might not be such a good choice as an interconnect between a carrier and a mezzanine.

Mezzanine or Mini-Blade?
What seems clear at this time is that Ethernet is the preferred transport protocol in most situations for most people, and its popularity just keeps growing. I have been a firm subscriber to the “Ethernet Vincit Omnia” ideology. As we have seen, few technologies can compete with Ethernet. Ethernet is familiar, affordable and available. It is the fabric of choice for AdvancedTCA and MicroTCA backplanes. However many of today’s systems are being designed with other topologies (namely PCI Express). This poses something of a dilemma for AdvancedMC designers, because Ethernet does not always make sense as the interconnect between an AdvancedMC and its carrier.

The fact is that many AdvancedTCA system designers prefer a carrier-to-mezzanine interconnect based on PCI Express. The idea is to centralize high-performance processing and provide low-cost, high-density I/O on AMC modules.

PCI Express currently delivers the best possible price/performance for high-density I/O applications. In contrast, some system designers prefer a distributed processing environment using Ethernet as a backplane and more costly “Intelligent I/O” AMCs using an embedded General Purpose or Network Processor architecture.

There are other fabrics like Serial Rapid I/O and SATA, each with their own specialized niche applications, but the overwhelming number of AMCs on the market today are based on PCI Express. This reality helps explain why SBS has chosen to design AdvancedMCs based on PCI Express.

AdvancedTCA and BladeCenterT systems will continue to use Ethernet and 10 Gigabit Ethernet or Xaui on the backplane. So too will the emerging MicroTCA specifications. That last point seems to worry people in the industry. If MicroTCA is based on Ethernet and the majority of available AMCs are PCI Express, do we need to design two AMCs for each application?

The MicroTCA system lends itself very well to a distributed processing architecture. Therefore, “Inteligent” AMCs like SBS’ Telum 1200 Network Processor based products are needed. We refer to these AMCs as Mini-Blades internally here at SBS. These AMCs cost more than their “non-intelligent” PCI Express based counterparts, but the cost of computing is spread across multiple AMCs in the system. I also believe that there will be low-cost, high-volume MicroTCA systems based on PCI Express in the future as well.

So the ultimate reality seems to be that although AdvancedMC cards can theoretically plug into the AdvancedTCA carrier as a mezzanine or the MictoTCA backplane as a mini-blade, they will most likely be designed primarily for only one of those scenarios. This does not undercut the goal of volume pricing based on total interchangeability because there will still be a tremendous overlap between the application needs and capabilities of the two types of systems.

.1) .2) .3) .4) or .5)–All of the above.
As the AdvancedTCA/MicroTCA ecosystem evolves, it will probably incorporate many different fabric implementations based on the needs of different applications. This has always been the case in the computer industry, even with different types of buses. For example, when PCI could no longer handle the bandwidth needs of PC graphics, AGP came along. Today, PCI Express is replacing the venerable PCI bus, however most desktop computers will probably support both technologies during the transition.

The AdvancedTCA/MicroTCA family will likely evolve along similar lines, taking advantage of the available fabrics when they make sense for any particular situation. So the question, “Which fabric is best?” really has no good answer. The best fabric is the one that produces the best results for your application, be it data, voice, video or storage. Each designer of each system will need to make these judgments, and thanks to the thorough work of the specifications committees, there are many good technologies to choose from. And of course, SBS is always glad to help you make your decisions.

By Jeff Marden, Technologist - Communications In the title and discussion below, the terms MicroTCATM Carrier Hub (MCH) are used instead of Virtual Carrier Manager (VCM) to describe controller functionality for the MicroTCA system. In one of the recent MicroTCA Subcommittee conference calls/meetings, the committee members voted to change the name of the entity that is used to control the MicroTCA system as described below.

AdvancedMC^TMs were originally designed to reside on AdvancedTCA® carrier cards. Only later were they proposed as mini-blades in MicroTCA^TM systems. Because these mezzanine cards were not originally designed to be blades, the MicroTCA specification writers had to devise a scheme to enable the MicroTCA backplane to emulate an AdvancedTCA carrier card.

One of the primary obstacles was that a backplane is generally rather “dumb” in the sense that it is little more than a set of electrical connections. Most of the intelligence comes from the blades inserted into the backplane, such as the switching and CPU blades in AdvancedTCA systems. Carrier cards and payload blades must also have enough intelligence to handle certain system functions such as power, clocks, system interconnections, as well as their own functional status including resource and thermal management.

AdvancedMC cards do not carry this level of capability, and so MicroTCA systems had to replace this intelligence without modification to the existing AdvancedMC. Remember that one of the primary goals of MicroTCA was to use AdvancedMC cards with absolutely no modifications. In other words, the very same cards that could plug into an AdvancedTCA blade had to be able to plug into a MicroTCA system. This meant that the functions of the AdvancedTCA carrier could not be added to the AdvancedMC.

The authors of the MicroTCA specification chose to solve this challenge by creating a “virtual carrier” to support the AdvancedMC cards. The idea was to transform the MicroTCA backplane into a carrier environment which could replicate the intelligence of both the AdvancedTCA carrier and switch blades. This allows the unmodified AdvancedMC cards to plug directly into the MicroTCA backplane as they would a carrier card. Of course, while an AdvancedTCA carrier can take just 4 single wide, full height AdvancedMCs, a 19” MicroTCA chassis can take up to 12 payload cards, so the MicroTCA Carrier Hub (MCH) is significantly more capable than an AdvancedTCA carrier.

For the “virtual carrier” concept to be successful, there was one essential element which still needed to be created, and that was the controller function; the MCH. This new type of AdvancedMC-like card would provide intelligence to the MicroTCA backplane. In many ways, this new card would essentially transfer the “intelligence” which resides on the AdvancedTCA carrier onto the MicroTCA system. This seemed a daunting task, given the board real estate available.

The task of developing these new cards has already been undertaken by several vendors. Fortunately, early prototypes which have been produced and tested in MicroTCA systems have succeeded in delivering all the functions needed to make this new specification a success. This is particularly encouraging considering the board size limitations of the AMC.0 form factor: 75mm x 181mm.

The MCH provides an Intelligent Platform Management Interface (IPMI), which is capable of handling the twelve Advanced Mezzanine Cards plus the MicroTCA chassis itself, including the backplane, cooling and power modules. In addition, the MCH manages the base fabric through a switch which connects the system cards, and can also be responsible for handling the data fabric which will be used for applications running on the system. Further, the MCH is responsible for managing the three distributed telecom clocks and power distribution. This is an amazing amount of capability to fit onto an extremely small amount of board space, but the prototypes have proven it can be done.

Without the MCH, every other card in a MicroTCA system is essentially useless, so it is quite likely that high availability systems will include a second, redundant MCH to eliminate a single point of failure. And, in the future, fairly extensive remote diagnostics will also be added to the MCH through JTAG testing capabilities.

As you can see, the MCH is critical to the operation of a MicroTCA system. Present and future suppliers will provide MCH products that will allow system designers to take advantage of the amazing promise of this new specification.