Monday, January 28, 2008

Portland’s Ethernet espresso

It’s still early days for the next big step in Ethernet networking, the leap to 40 and 100G standards. But the newly approved IEEE P802.3ba task force is on the case and hopes to have a cogent draft by this fall.

A look at the group’s agenda from Portland last week shows they have been drinking double espresso. The agenda included more than 30 presentations from a wide variety of players including top dogs such as Alcatel, Broadcom, Cisco—the ABCs of Ethernet.

Interestingly, HuaWei is stepping up to take a significant role in the committee and is making solid technical proposals. The days of hacking on Cisco code are over for this emerging powerhouse.

Defining a common electrical interface has become one early issue for the standard. Joel Goergen of Force 10 Networks did a good job advocating for using the practical 4x10G and 10x10G links initially, moving to 4x25G in future—and maybe some 10x3 or 8x6 action in the medium term. Leveraging the 10GBase-KR standard for serial 10G links and getting in step with the Optical Internetworking Forum’s efforts on 25G will be key, according to Goergen.

If there are broader issues that need to be aired here, I am all ears. Post a note or drop a line to rbmerrit@cmp.com. In the meantime...Man, I am going to have get me some of that good Portland coffee!

3 comments:

Anonymous said...

Seeing all the hoops that people have to jump through now to run 10G on copper, does it even make economical sense to flog further? I assume that's what you're talking about in "common electrical interface".

I would think that fiber technology and installation costs would have dropped for these speed applications to retire copper once and for all. (Not that copper hasn't served us well for everything in the past, but let's move forward, no?)

Miriam from McBru said...

Hey, I'm going to Byways for breakfast on Thursday!

Anonymous said...

10GbE KR does just fine on the backplane and with the rise in blade adoption eliminate the need for cable waterfalls and complexity. As shown in prior IEEE meetings, 85% of the cable runs are relatively short enabling copper cable solutions to be constructed at much lower power points and cost. For those cable runs that require longer distance, active cables from several vendors can provide a lower power solution.

So the question isn't copper vs. optics but why aren't vendors developing products that target the specific use models and avoiding the pitfalls found by adhering to the worst-case scenarios defined in standards?

 
interconnects