AMC Insider - August 2006

Gene Juknevicius Technologist, GE Fanuc Embedded Systems AdvancedTCA, AdvancedMCTM and the recently ratified MicroTCA specification are built around serial interconnect technologies such as PCI ExpressTM, Ethernet, SATA, SAS, InfiniBand® Technology and Serial Rapid IO. Even though Ethernet is the most ubiquitous

1000BASE-BX Overview
Driven by design challenges, the PICMG committee defined two PHY options: 1000BASE-BX and 10GBASE-BX4. 1000BASE-BX uses two differential pairs operating at 1.25 GBd, one for transmit and one for receive. 10GBASE-BX4 uses eight differential pairs (four for transmit and four for receive), each operating at 3.125 GBd. Both 1000BASE-BX and 10GBASE-BX4 are used in AdvancedTCA, AdvancedMC and MicroTCA as a Physical Layer Interface (PHY).

The 1000BASE-BX physical layer was adopted from the 1000BASE-X definition. Figure 1 shows the Ethernet layers and indicates the origins of 1000BASE-BX. 1000BASE-BX, as defined in the PICMG 3.1 specification, was in fact a SerDes interface used to connect the Physical Medium Attachment (PMA) with fiber optic transceivers.

Figure 1 SerDes Location in Ethernet Sublayer Stack

Ethernet Autonegotiation Function
Ethernet Autonegotiation is defined in the IEEE 802.3 specification. Fundamentally, Autonegotiation enables dynamic selection of otherwise conflicting features. At the time the PICMG 3.1 specification was released, the IEEE made no official effort to define Ethernet over backplanes and applicable Autonegotiation. Nevertheless, PICMG by adopting the 1000BASE-X definition and adding additional electrical requirements, also adopted Autonegotiation as it is defined for 1000BASE-X. 1000BASE-X Autonegotiation allows devices to dynamically select the following options:

• Duplex (half duplex, full duplex)
• Flow control (symmetric pause, asymmetric pause, no flow control)
• Indicate partners failure
• Exchange status information

Different media use different messaging methods to exchange Autonegotiation information. 1000BASE-X exchanges Autonegotiation information in-band using /C/ ordered sets defined for 8B/10B encoding. Since 1000BASE-BX supports only 1Gb/s data rate speed, negotiation is not applicable. Other abilities, however, could be exchanged with a link partner and negotiated upon. During Autonegotiation, two interconnected devices exchange ability records then, based on the priority scheme, select the highest common denominator.

Recommended Autonegotiation Settings
The IEEE specification defines provisions to disable Autonegotiation and to manually configure link parameters such as speed, duplex and flow control. This is the area where most interoperability issues occur.

It is important to note that if one device disables Autonegotiation, then its link partner must disable it as well. Disabling Autonegotiation on only one side most likely will prevent the link from being established.

Since the PICMG specification does not give guidance as to whether or not 1000BASE-X Autonegotiation should be disabled, vendors are free to do as they please. If Autonegotiation is to be disabled, at the very least, manual link settings such as flow control should be agreed upon. Manual settings are not captured in the current standard and the IEEE specification recommends keeping Autonegotiation enabled.

Autonegotiation provides a more elegant way to achieve the desired link configuration than using brute force. It allows you to control which features are being advertised to communication partner. If, for instance, a standard device is used for 1000BASE-BX, the device’s PHY may support multiple speed options. Since only 1000Mb/s speed is applicable to 1000BASE-BX, the other speed options should be disabled in the Autonegotiation advertisement register. Doing so allows Autonegotiation to be used, but ensures that only the supported mode of operation will be negotiated to. More specifically for 1000BASE-BX which is used in AdvancedTCA, AdvancedMC and MicroTCA, the following parameters should be enabled in the Autonegotiation advertisement register:

• Only 1000 Mb/s speed
• Only full duplex operation
• Supported flow control options

One final note—at the time of this writing, definition of the IEEE 802.3ap specification is in progress. IEEE 802.3ap defines 1Gb/s and 10Gb/s Ethernet over backplanes as well as Autonegotiation between the two. Such Autonegotiation is not discussed here and will require PICMG efforts to adapt it, since it directly conflicts with E-Keying functionality.

References
IEEE 802.3-2002, Carrier Sense Multiple Access with Collision Detection (CSMA/CD) access method and physical layer specifications

IEEE 802.3ae-2002, Carrier Sense Multiple Access with Collision Detection (CSMA/CD) access method and physical layer specifications. Amendment: Media Access Control (MAC) Parameters, Physical Layers, and Management Parameters for 10 Gb/s Operation

PICMG 3.1, Ethernet/Fibre Channel Over PICMG 3.0 Specification

Rubin Dhillon V.P. Communications & Enterprise A couple years ago we decided to embark on an aggressive strategy to increase our market share in the Telecommunications space. Our whole strategy was based around the idea that AdvancedMC products would be wildly successful in the architecture of Modular Communications Systems platforms.

It worked.

We have managed to take a leadership position in developing AdvancedMC products and related technologies and we have more engagements with customers, partners and technology providers than ever before. However, there has been one big surprise for us – the amount of interest coming from NON TELECOM customers.

One of the unique advantages of being a company with name recognition in multiple markets is that we have always managed to get attention from Military and Industrial/Commercial customers looking for Communications Technology. We introduce technologies for one target market that hops over to another vertical market. A good example is our Ethernet Switch technology that was introduced to the Communications market some years back and is now very successful in rugged Military systems.

We always intended to introduce AMC technology to non-Telecom applications, however, the immediate high interest level in AMCs from all market segments took us by surprise.

So why do I think this has happened?

Reason #1: Modular Computing Architectures are applicable in all markets and the AdvancedMC card is a nearly perfect expression of modularity. The AdvancedMC offers a high-performance, highly-available technology in a compact mezzanine form factor. It offers a technology re-use and refresh architecture with manageability down to the mezzanine card. It has a growing vendor base, offering competitive, cost effective solutions. The card is so modular that it functions today in AdvancedTCA® carriers, IBM® Blade Center systems, other proprietary systems, and even as a blade in MicroTCA^TM systems.

Reason #2: The architecture is very flexible. AdvancedMC cards may support a wide range of compute and I/O functions as well as a wide array of fabric interconnects. AdvancedMCs are being deployed with PCI Express^TM, Gigabit Ethernet, Serial Rapid I/O and 10 Gigabit Ethernet interconnects today. It scales up to AdvancedTCA and down to the MicroTCA^TM pico chassis. The AdvancedMC real estate is large enough to support the dense computing platforms, and high bandwidth I/O used in today’s demanding applications like IMS and IPTV infrastructure devices.

Reason #3: MicroTCA. This is probably going to be the most important single factor behind the success of the AdvancedMC card. This is where I see a great deal of potential in bringing the AdvancedMC architecture to applications beyond Telecom. It is entirely feasible to see the system architecture deployed in enterprise, medical, military and industrial applications. The MicroTCA architecture is gaining interest amongst our Military customers as it offers high-performance in a small form-factor. Our Enterprise and Industrial customers are intrigued by the promise of cost effective system building blocks

Reason #4: Momentum. One of the greatest hurdles for a new technology is getting it widely accepted. Based on our engagement with customers, I believe this hurdle has been overcome. Sure there have been some concerns with maturity as there have been various discussions about specification changes both benign and radical. There have been some initial growing pains with interoperability and compatibility. But it seems we have most of these concerns behind us now and companies such as GE Fanuc are committed to making it work.

It has been quite some time since the AdvancedTCA standard was finalized. Only now, years later, is real equipment being deployed. The AdvancedMC specification was ratified about a year ago, and it immediately saw strong demand. MicroTCA, which hasn’t even been finalized by the subcommittee, is generating an almost unbelievable amount of interest. Which is exactly what you’d expect: the whole ecosystem is gathering momentum, and once that happens more and more people start jumping on the bandwagon.

Which brings me back to the title of this article. If you haven’t given this modular, open technology some serious consideration, now is the time, because the bandwagon is picking up speed. There’s still plenty of room for you, so get aboard. I predict it’s going to be a wild ride.

Jeff Marden Engineering Fellow MicroTCATM, the most highly anticipated embedded specifications to be released in a decade, has been unanimously ratified, and has been published. This is very exciting news for those of us on the subcommittees and working groups who have labored long and hard to get the details hammered out.

More importantly, this is a significant breakthrough for the entire embedded computing community. MicroTCA takes what has been learned from AdvancedTCA® and delivers it in a small and yet incredibly potent package. Because of this, I believe MicroTCA will spread from communications applications into the rest of the embedded market space.

There have been a lot of rumors about exactly what was included in the final specification, and there may be a significant amount of confusion. So in the next few paragraphs I would like to give a short, birds-eye view of what I consider to be the most significant features of MicroTCA.

MicroTCA Carrier Hub
AdvancedMC “min-blades” were originally designed to plug into ATCA carriers, so MicroTCA systems have to replicate that environment. However, once that requirement is met, the specification allows quite a bit of leeway because there a number of different potential interconnects, and there is even the possibility that for smaller, simpler systems the carrier functions will be placed on the backplane.

Full AdvancedMC Module Support
As per the Statement of Work for the MicroTCA specification development, AdvancedMC Module are supported without modification to the Module form, fit or function.

Card Sizes
The original AMC.0 form-factors have been incorporated into the MicroTCA specification, however no AMC.0 ECN changes have been addressed. This means MicroTCA supports four different mechanical sizes of cards which can be slotted into a system: Full- and Half-height, Single and Double widths, and there are provisions for mixing these four different sizes within one system.

Chassis Sizes
The range of potential system sizes here is pretty amazing. Everything from “pico” which is a couple of cards, a power supply and a fan, to full sized, rack-mountable chassis. This is another exciting feature of MicroTCA: the fact that it can scale so far up and so far down in size and capability.

Management Features
MicroTCA offers all the management capabilities that are part of AdvancedTCA because, once again, the systems are based on that foundation. These management capabilities include low-level hardware management services based on Intelligent Platform Management (IPMI) and high-speed management services based on IP protocol suites.

Scalable Availability
In some instances, MicroTCA units will be placed near the edge, or at the very edge of the network. Because of this, they do not always require 99.999% availability. The loss, for example, of a single WiMAX base station at a network edge is probably not a disaster. So MicroTCA allows for differing implementations of redundancy in power supply, cooling, management and switching based on the location of the equipment in relation to the network core.

Now the fun begins
If the history of technological innovation teaches us anything it is the simple fact that the “best” technology does not always win—VHS vs. Beta being the most oft-cited example. There’s just no way to predict what will happen with new technologies, so it will be a lot of fun to see how MicroTCA fares in the next few years.

From a purely technological point of view, MicroTCA certainly looks like a winner but it will have to pass a number of tests before we declare it a success. Will it achieve high enough production levels to meet its pricing goals? Will a strong vendor base develop? Will it find acceptance outside the telecom market? Only time will tell, and now that the specification is ratified, the clock is ticking.