Optical components

Backplane Bandits Get Together

Next week, a group calling itself the High Speed Backplane Initiative (HSBI) is planning to launch.

The group's goal is to promote a common solution for the next hike in backplane speed, the primary conduit of bandwidth in many pieces of high-powered routing and switching gear. They see a need in the next year or so for backplanes that can support serial electrical signals at 5 Gbit/s -- twice the speed of today's state-of-the-art 2.5-Gbit/s backplanes.

The need is being driven by the fact that optical modules are getting smaller and line rates are going up, says John D'Ambrosia, secretary of HSBI and manager of semiconductor relations for Tyco Electronics. Systems vendors are talking about putting four or eight 10-gig modules on a line card, and at that density there are so many traces on the backplane that it becomes impossible to route them all.

A twofold increase in backplane speed might seem a little strange when the electronics of telecommunications interfaces are increasing by jumps of four (Sonet) or ten (Ethernet) times. But trying to achieve a fourfold hike in backplane speed didn't look technologically possible within the timeframe it was required, says Bill Woodruff, VP of marketing for Velio Communications Inc., another founder member of the HBSI.

In fact, work has only just begun on solving the technical problems of sending 5 Gbit/s over 30 inches of copper. "That's why we got together, 'cause we said, 'This ain't easy,' " Woodruff quips. There are a variety of ways of solving the technical issues, he says, and left to their own devices, components vendors would likely come up with an array of incompatible solutions.

Interoperability is vital, says Joel Goergen, chief scientist at Force10 Networks Inc., especially for a commodity part like a SerDes transceiver -- the chip that sends and receives the high-speed electrical signal over the backplane. Being able to chose compatible parts from a range of suppliers is essential if vendors like Force10 are to build next-generation systems cost-effectively.

There seems to be a broad consensus on that. The HSBI has nine founding members, mostly components vendors, including Tyco, Velio, Agilent Technologies Inc. (NYSE: A), Intel Corp. (Nasdaq: INTC), Gennum Corp. (Toronto: GND), Marvell Technology Group Ltd. (Nasdaq: MRVL), Mindspeed Technologies, Tality Corp., and Texas Instruments Inc. (NYSE: TXN). Systems vendors are represented among the 28 "supporter" companies that have been involved in discussions but have not signed a legal agreement.

The group has already made a couple of decisions. One of its early conclusions was to support multiple coding protocols over the backplane. These protocols decide how multiple 2.5-Gbit/s Sonet data streams are combined and split apart, or how a 10-gig data stream is broken up and recombined, so any size channel can be sent over a 5-Gbit/s link. In general, different schemes operate for Sonet- and Ethernet-based traffic.

"If some member voted for one scheme and some for another, then it's likely that we'd have a fragmented group," says Tyco's D'Ambrosia. "By making peace with everyone, we get more expertise -- and more market penetration."

The other key question was whether to use a modulation scheme or to use non-return-to-zero (NRZ) signaling. The group picked NRZ. "We believe we can do this more cost effectively with NRZ," D'Ambrosia contends. Virtually all backplane transceivers today use NRZ, so vendors have plenty of experience with the technology. And it is easier to integrate with VLSI integrated circuits, which use similar signaling techniques internally.

However, that decision could leave startup Accelerant Networks Inc. out in the cold. Accelerant is the only company with a 5-Gbit/s backplane product to date, but it uses a multilevel modulation scheme (see Copper Battles Optics in Backplanes and Accelerant Boosts Backplanes).

HSBI is still in a very early stage of definition, says Jim Tavacoli, Accelerant's VP of marketing, and there is not enough concrete definition available for anybody to assess whether such an approach is going to work or not. Whereas Accelerant's solution, he claims, "has proven performance at up to 6.25 Gb/s on over 50 existing backplanes, some of which were designed to operate at speeds as low as 622 Mhz."

Accelerant is a supporting member of HSBI, he adds, and continues to stay abreast of new developments.

— Pauline Rigby, Senior Editor, Light Reading
http://www.lightreading.com Want to know more? The big cheeses of the optical networking industry will be discussing this very topic at Opticon 2002, Light Reading’s annual conference, being held in San Jose, California, August 19-22. Check it out at Opticon 2002.

Register now and save $500 off the registration fee. Just use the VIP Code C2PT1LHT on your registration form, and deduct $500 from the published conference fee. It's that simple!

Page 1 / 3   >   >>
52395 12/4/2012 | 10:02:27 PM
re: Backplane Bandits Get Together "They see a need in the next year or so for backplanes that can support serial electrical signals at 5 Gbit/s -- twice the speed of today's state-of-the-art 2.5-Gbit/s backplanes... Systems vendors are talking about putting four or eight 10-gig modules on a line card"


Umm, which systems vendors are in a sufficient state of denial to ponder 4-port (or 8-port!!!) 10Gig line cards in the next 2-3 years?

If I'm a carrier, my DS-3 revenue is ~$5K/month and I believe in Ethernet for the WAN, then 1 10Gig port will carry $1.1 million/month. For redundancy I'll want separate paths and probably separate routers, but at the very least separate line cards for that kind of revenue. Maybe I'd use a 2-port card to have multiple cards back each other up. But 4-ports on one line card? An operational disaster waiting to happen.

If I'm a competitive carrier with big kahunas (are there any of those left?), maybe I price 100Meg access for ~$1000/month. Now we're only talking about $100K revenue/month/port. I'm probably also expecting data-only with a good stat-mux gain so I probably oversubscribe the 10Gig links, but let's assume it's just $100K/month. Same argument above applies.

If I'm an enterprise, sheesh, 40-80 Gigabits/second on one line card? Where's the need???

- P
hitekeng 12/4/2012 | 10:02:25 PM
re: Backplane Bandits Get Together There might be a need for mutliple ports 10Gig card in a couple of years from now. Who would use it!!! Cogent for instance, who had been disruptively offering 100Mbps @ $1000, has recently acquired some of (I do not recall if it was MFN's) a bankrupt carrier dark fiber assets to offer the service of linking tier 1 metros together (as competition in that area is dwindling with the demise of carriers such Sigma Networks and so on...). So a single card entails less spares and OAM&P leading to less cost overall..
next-gen-wisdom 12/4/2012 | 10:02:22 PM
re: Backplane Bandits Get Together With a multiple GigE ports card in your Sonet eqpt. you can interconnect your edge rings to your different GigE Ethernet switches in the POP and distribute the 10/100BaseT traffic thru your different MANs or COs.

This make sense?
raypeso 12/4/2012 | 10:02:22 PM
re: Backplane Bandits Get Together I guess it's a natural progress but I can't see to large of a market for it. Maybe for equipment in very large buildings, but it seems kind of difficult to find a place where you could unload 8 10gig circuits. I work for a CLEC and we almost never find that kind of DWDM penetration in one spot.
sgan201 12/4/2012 | 10:02:21 PM
re: Backplane Bandits Get Together Hi,
There is a DWDM box that can take in 4 X 1310nm and send it 20 KM without electrical conversion for $30K each. It supports up OC-192/10Gbps without caring whether it is ESCON, Sonet or 10GE..
I do not think any electrical solution can beat this price..
boozoo 12/4/2012 | 10:02:20 PM
re: Backplane Bandits Get Together You sound a little bit like Bill Gates.
Just because there's no demand now does not mean demand will not be here in 5 years.

DrFager 12/4/2012 | 10:02:19 PM
re: Backplane Bandits Get Together If you're interest in some of the shenanigans that caused the collapse of the telecom industry, then the cover story on the Business Week website is very informative. The story can be found at:

jodam 12/4/2012 | 10:02:18 PM
re: Backplane Bandits Get Together Requirements for backplane speed are not just driven by the I/O. Backplane architectures play a driving role as well. For example, you can design more bandwidth into the backplane in order to provide better quality of service.

The need may not be evident based on the reasoning you are providing..... but it is there....
sgan201 12/4/2012 | 10:02:16 PM
re: Backplane Bandits Get Together Hi,
There is always a demand for something..
However, there might be market for only one box.
The difference between a science project and profit making product is

1) Science project try to push the boundary of technology without thinking about whether the market will be big enough for the end result to be paid off..

2) A real profitable product is created only when the market is big enough to buy/pay for the end product profitably.

The difference between 1 and 2 is timing..
1 is ahead of time..
2 is just in time technology..

Based on what I said and seen, eletrical switching have to be very cheap to beat the optical DWDM solution.. What they will come out will not be cheap enough since it is leading edge technology. Hence, it will not have a large market. Then, it will not be profitable.
In summary, it is a science project...
WolfLarsen 12/4/2012 | 10:02:15 PM
re: Backplane Bandits Get Together
Does anyone know how this effort flows into the advanced telecom computing architecture (aTCA) being worked by PICMG?

Page 1 / 3   >   >>
Sign In