Startups Move 10-Gig to Copper
The "10-G Base T" study group hasn't set a timetable yet, but initial backers expect to complete the standard within the four-year window allotted by the IEEE. The group will be part of the 802.3 series of standards for Ethernet.
SolarFlare's pitch was backed by five other startups: Cicada Semiconductor Inc., Accelerant Networks Inc., Mysticom Ltd., Plato Labs, and Telicos. This clique won't necessarily be running the study group that the IEEE approved; those kinds of administrative details will be ironed out in the coming months.
"Our worst issue was the cynics," says Ahmet Tuncay, SolarFlare's founder and vice president of marketing. "Most people thought 10 Gbit/s could not be brought onto this type of cabling because it was beyond the Shannon wall" -- that is, beyond the theoretical limit of transmission speeds on a noisy channel.
SolarFlare's presentation -- which went into significant technical detail -- convinced the engineers in attendance that the feat was possible. "You just have to mitigate more of the capacity-reducing impairments like echo," Tuncay says.
The IEEE already addresses 10-Gbit/s Ethernet in its 802.3ae standard, but the new group's focus is on a deeper level. Specifically, they'll be discussing the electronics required to make 10-Gbit/s signals viable on existing copper wiring, as the noise levels on those cables can be enough to slow down or disrupt high-speed signals.
The goal is to make 10-Gbit/s Ethernet viable for the existing business environment without forcing companies to rewire their buildings with fiber.
"Unlike metro Ethernet, enterprise Ethernet has a legacy of working on installed cabling, and UTP [unshielded twisted pair] is the predominant flavor of cabling in these data centers," Tuncay says. "The real challenge of 10 Gbit/s in the enterprise has been to figure out how to make it run on this installed cabling."
It's worth noting that the six startups aren't in lockstep agreement on the standard. They all want to use technologies available today, such as pulse-amplitude modulation and adaptive equalization, but the specifics are up for grabs.
PAM, for example, is a way for electronic signals to carry more than one bit per pulse, but the startups disagree on which version of PAM should be used. With adaptive equalization -- the ability to compensate for the variations in signal quality -- it's a question of whether an analog or digital technique is best.
"These are the areas where the study group is going to spend some time," says Jim Tavacoli, vice president of marketing for Accelerant.
The study group's next step is to submit a Project Authorization Request to the IEEE. Once that's approved, the group becomes a "working group" and is officially on its way toward defining a standard.
— Craig Matsumoto, Senior Editor, Light Reading
www.lightreading.com
Want to know more? The big cheeses of the optical networking industry will be discussing 10-Gigabit Ethernet at LightSpeed Europe. Check it out at http://www.lightspeedeurope.com.
"to make 10-Gbit/s Ethernet viable for the existing business environment without forcing companies to rewire their buildings with fiber."
would be an easier to sell if there was some historical market data on the re-use of Ethernet wiring that took place going from 10Base-T to 100Base-T or 100Base-T to 1000Base-T. Otherwise I can't judge if the goal is worthwhile or just hype.
I can defintely see the value of 10,000Base-T on a new CAT cabling category with an appropriate connector for short distance , say less than 100m.
A less critizable, but la ess-laudable goal would have been to have it be able to auto-sense between 1000Base-T and 10,000Base-T.
Also the startups may be under-estimating the effort involved here. This is not a trival effort, irregardless of their laudable goal of cable re-use, based on my experience trying to get 10Gbps electrical over backplanes. But hey, it will happen given enough resources.
For some of the difficulties they will be dealing with see "Nov 21, 2002: Proper Serdes Selection Solves Serial Backplane Design Woes" http://www.commsdesign.com/sto...
for some public information on this topic from PMC-Sierra.
This is also in keeping with recent work by the OIF Physical Link Layer (PLL) Working Group announcing "a new project aimed at defining a new electrical I/O specification for 5- and 10-Gbit signaling. According to the OIF, the working group will work on defining specs for short reach (up to 200 mm with one connector) and long reach (up to 1 meter with up to two connectors)". See "Nov 21, 2002: OIF specs address optical packaging, performance " http://www.commsdesign.com/sto...
MrLight :-) Fully supporting the 10,000Base-T effort since it will help with 10Gbps over the backplane.