The Great Gigabit Backplane Shootout - Question
#2
There has never been a standard for backplane transceivers. The closest example of a standard has been XAUI, which is an Ethernet standard that has been co-opted for the backplane. However, in spite of its broad acceptance, finding XAUI products from two different vendors that will reliably interoperate is more than challenging, and often impossible.
The primary reason for the lack of standardization of backplane transceivers has been the fact that both ends of the transmission line are contained in the same equipment. Since the transmitter and receiver are controlled by the same entity (the system manufacturer), there is no need for a standard to ensure operation with an unknown vendor. In addition, the backplane has been a key area of differentiation amongst communication equipment vendors. For many, their backplane design is a closely held secret providing a competitive advantage that enables them to create more efficient, reliable and higher bandwidth backplanes. This in turn allows equipment manufacturers to create systems that offer the optimal balance of port densities, number of line cards, port speeds and services offered.
Looking forward to 6.25Gb/s and higher, it will be necessary for the backplane transceiver to do much more than it did before. By increasing the transmission frequency, problems far greater than those previously encountered are exposed, requiring innovative and targeted solutions. These solutions will vary widely from company to company based on their past experiences, core competencies and a belief in particular solutions. While it may be possible to establish a standard, having transceivers from different vendors that reliably interoperate will continue to be a significant challenge. Without true interoperability the main purpose of a standard is defeated.
Finally, there will always be an unrelenting demand for market leading products. Those who offer the best performance, lowest bit error rate, and best features will always be in great demand.
Most equipment providers require multiple vendors for the components
in the system. Furthermore, the devices on the other end of the backplane
could be provided by a competitor. For these reasons, interoperability is
required for SERDES products.
Interoperability is either achieved through industry standards or proprietary
standards driven by equipment providers.
There are two significant activities underway today: OIF and HSBI.
Although the OIF has just started looking at higher backplane speeds, they have tentatively selected 10Gbps as their future standard. OIF 2.5Gbps solutions currently include plans for 40Gbps systems (SPI5, SFI5, TFI5). Since the market for 40Gbps systems has been delayed, it is not unreasonable to assume that the OIF will create a 10Gbps standard for a 40Gpbs system. The precedent for this was set with the development of SFI4 phase 2.
HSBI is investigating 5Gbps with a future path to 10Gbps. At this time the membership includes backplane/connector and IC suppliers. The equipment vendors will need to either join this group or embrace their recommendations before it will be recognized as a standard.
The market will dictate useful results from the standards bodies. If the market has a significant upturn, the system vendors will require innovative solutions, and the silicon vendors will provide them. The system vendor(s) who gain the most traction in the market earliest will sway the standards bodies.
Interoperability is of paramount importance. It reduces the risks associated with new technology. They enable competition and lower the cost of the components. There are several groups that are working in this effort, here are a few of them:
Players:
For higher speed SerDes devices, standards such as the IEEE802.3ae have been developed to ensure interoperability between suppliers over XAUI based systems. While some system manufacturers use multiple SerDes vendors over their backplane and interoperability is an issue, there are also a number of proprietary backplanes where only a single source of SerDes is required. This should hold true for higher speed backplanes as well, where groups such as the High Speed Backplane Initiative (HSBI) are being formed to address interoperability issues. Proprietary backplanes will most likely remain that way even as they move to the higher speeds.
From an equipment vendor's viewpoint, component standards are a good thing. They spur competition and drive down pricing while lowering manufacturing risk through multiple sources of supply. However, standards must solve a customer's problem before they become useful. To do this, customers must be actively engaged in the standards development. Today there are four camps of backplane SERDES and transceiver vendors using widely varied technologies with various levels of engagement with customers.
The first group is a large consortium of component companies grouped together as HSBI (High-Speed Backplane Initiative, www.hsbi.org) who look to extend the life of two level signaling. They plan to do this by increasing the signaling rate in the range of 4.976 Gb/s to 6.375 Gb/s. This option is attractive to many at first glance since it piggybacks on much of the two-level work done in the past. Much of today's equipment is utilizing 2.5 Gb/s or 3.125Gb/s two level SERDES so the HSBI solution would be a 2 - 3X increase in signaling speed over what they are using today. The group indicates that it hopes to complete their specification by the end of the year and have member companies supply samples in 2003.
The second group includes the various companies supporting simplex multi-level signaling. This includes companies like Accelerant Networks using 4 PAM technology and MathStar using 5 PAM technology who both are sampling devices today to customers for evaluation. There are also several other companies developing simplex multi-level parts for sampling in 2003. While these companies are all using multi-level signaling to maintain or slightly lower today's signaling rates, there is no common standard for these parts to inter-operate.
Keyeye Communications in conjunction with another IC vendor has created a third group utilizing 4 PAM full duplex technology. With the major system performance limitations today being crosstalk in and around the connectors and power (not attenuation), keeping the signaling speed well below 2G is critical. KeyEye's "EchoWave" technology, makes use of a 4PAM full duplex architecture in conjunction with adaptive receive equalization to offer true "plug and play" performance in the backplane. Legacy mode support for 1.25G / 3.125G SERDES is designed into our devices to help support installed equipment.
The last group is being driven directly by large OEM's. These companies whose volume can justify IC vendors delivering proprietary solutions solely to them are actively involved in this space. Various types of two-level and multi-level technology are now being developed for customers with high-speed samples expected late this year and in 2003.
Standards are critical for widespread adoption of backplane technology. At 3.125 Gbps data rates, XAUI is the de-facto standard. At 5 to 6 Gbps, HSBI is leading the charge. HSBI is likely to have a preliminary specification released by end of the year.
Very little today. The only efforts that were successful were driven by the 10 Gigabit Ethernet Alliance through the interoperability working group. This committee was formed to validate the feasibility of XAUI and was not intended as a performance benchmark but as a verification of compliance. All results are confidential and some preliminary information was disclosed at IEEE 802.3ae meetings. Some SerDes vendors that participated in related interoperability tests generated press releases to confirm their compliance to the standard. Mindspeed was one of these companies.
OIF will likely be a very relevant standards body in the SERDES backplane area. Standards are important in establishing interoperability between vendors, reference backplane models, common terminology and test methodology.
Standards are critical to insure mulitvendor interoperability. This is a key objective of the High Speed Backplane Initiative (HSBI). For more information on the initiative, visit www.hsbi.org. There is more than one way to achieve 5G or 10G performance over a bandwidth limited channel. A systems OEM should not be forced to use the same vendor on the switch card and the line card, and should instead be able to reply on industry wide, standards based solutions.