x
100G Ethernet

40 GigE Could End Standards Spat

A debate between server and switch makers in the Institute of Electrical and Electronics Engineers Inc. (IEEE) Higher Speed Study Group (HSSG) could lead to the addition of 40-Gigabit Ethernet to the next Ethernet standard.

Members of the HSSG will meet this week to discuss a Project Authorization Request (PAR) that is to be submitted to the IEEE executive council in November. But it's not clear whether the PAR will be for a 100-Gbit/s standard only, or one with the 40-Gbit/s rate also attached.

After the HSSG was founded last summer and the group voted to begin work on 100-Gigabit Ethernet, a standard seemed well on its way. (See 100-Gig Ethernet Takes First Step, 100-GigE Takes Shape, 100-Gig Ethernet Gets Official Nod, and IEEE Picks 100 GigE.) But in January, some members of the HSSG became more vocal in support of 40-Gbit/s Ethernet being added to the group's charter.

Server manufacturers such as Sun Microsystems Inc. , HP Inc. (NYSE: HPQ), Brocade Communications Systems Inc. (Nasdaq: BRCD), and others, led this group's demand for a 40-Gbit/s path that could be used for server interconnects. As a result, these vendors want a joint 40-Gbit/s and 100-Gbit/s proposal.

One supporter of the 40-Gig standard, who spoke on the condition of anonymity, says a combined standard is necessary because of problems with the current 10-Gigabit Ethernet standard. "The IEEE came out with 10-Gigabit Ethernet and it was adopted poorly because it was envisioned poorly."

He adds that "it will be 2015 before 100-Gigabit Ethernet gets any sort of volume." So that's why he sees 40-Gig Ethernet as a more economical alternative in the short term.

Jim Zik, senior product marketing manager for optical transport at Ciena Corp. (NYSE: CIEN), says many switching vendors think a 40-Gigabit Ethernet standard would delay commercialization of 100-Gigabit Ethernet for what many switching guys view as a niche application in the server-to-server market.

Niall Robinson, VP of product marketing at Mintera Corp. , says switching vendors would also be unhappy with the addition of cost and complexity to their products. "They would need to develop both interfaces on their equipment," he says. In addition to a 100-Gbit/s interface, he says "now they will also have to develop a 40-Gbit/s port, which will double costs" and increase development time.

While Ciena and Mintera are interested in a 100-Gigabit standard for transport, Zik and Robinson both said their companies were agnostic with regard to the outcome of the debate in the HSSG.

These differences have resulted in a divide between server and switching vendors in the study group. Votes within the group require 75 percent of members to vote in favor of a resolution to be passed. While the server camp is in a minority within the group, it has enough members to block a vote on any proposed standard that doesn't have 40-Gigabit Ethernet attached.

Many believe this has already caused a delay in the standards process and could threaten the future of the group itself if a resolution is not reached on the issue by the end of this week's meeting.

After a meeting in May failed to produce consensus on a PAR to be produced at the July plenary, HSSG chair and Force10 Networks Inc. components scientist John D'Ambrosia wrote in the group email that the group was in "a precarious situation."

"With the failure to get a PAR / 5 criteria approved at last week's interim meeting, we must now worry about gaining approvals for extension of the study group life," D'Ambrosia wrote. He also noted that it was "unprecedented" for a study group to get four extensions from the IEEE executive committee.

"This is about two markets growing at different rates trying to come to terms with one another," D'Ambrosia says. Nevertheless, he believes the debates "have helped to educate these groups about each other," and is optimistic that the HSSG will "reach consensus to move forward on a standard with two speeds."

That seems more and more likely, especially as 100-Gigabit Ethernet proponents move to support a dual standard. In an email to Light Reading last week, a Cisco Systems Inc. (Nasdaq: CSCO) spokesman wrote that despite its earlier support of a 100-Gig-only standard, the company would now back a joint proposal.

"Recognizing that the computing and server market and the networking and aggregation market have different needs, Cisco is amenable to and involved in discussion of a specification that supports two distinct rates of 40G and 100G," the spokesman wrote.

— Ryan Lawler, Reporter, Light Reading

Page 1 / 3   >   >>
Stevery 12/5/2012 | 3:05:21 PM
re: 40 GigE Could End Standards Spat One supporter of the 40-Gig standard, who spoke on the condition of anonymity, says a combined standard is necessary because of problems with the current 10-Gigabit Ethernet standard. "The IEEE came out with 10-Gigabit Ethernet and it was adopted poorly because it was envisioned poorly."

And so the solution to avoiding the 10GE fiasco is to make the next document MORE complex by incorporating 2 speeds.

Veeerrrry shrewd.
t.bogataj 12/5/2012 | 3:05:20 PM
re: 40 GigE Could End Standards Spat "In all the world, there is only room for a few computing machines."

"No one will ever need more than 640 kB of RAM."

"Who needs more than the dazzling 2400 bps?"

Sound familiar? What did market analysts say back then? Who will ever need 100 Gpbs?

Build it, and they will come.

tata, T.
Off_the_shelf 12/5/2012 | 3:05:20 PM
re: 40 GigE Could End Standards Spat The new Road to 100G Alliance (www.roadto100g.org) seems to have put their focus on what the market demands. That seems to make sense here too.

Maybe more people should be looking into this alliance and advocating this market approach. After all, 40G is on the road to 100G...
opticalwatcher 12/5/2012 | 3:05:20 PM
re: 40 GigE Could End Standards Spat "many switching vendors think a 40-Gigabit Ethernet standard would delay commercialization of 100-Gigabit Ethernet"

...only if 40 Gigabit makes more economic sense than 100 Gibabit Ethernet. Why not let the market decide if 40 or 100 Gigabit really makes sense?

Why was the 100 Gigabit speed chosen anyway? Was some tradeoff study done? Is it affordable? Does it make sense when looking at the end-to-end network and all its components? Or was it chosen, as I suspect, simply because it is 10x 10Gigabit?
tsat 12/5/2012 | 3:05:19 PM
re: 40 GigE Could End Standards Spat 40G is a reality today, in both optics and packet processing. 100G is still a long way off. If the market wants faster Ethernet quickly, 40G makes a lot of sense.

Otherwise, the world will be stuck with 10G Ethernet for several more years. It will only prolong SONET's lifespan.

-tsat
numinary 12/5/2012 | 3:05:18 PM
re: 40 GigE Could End Standards Spat There is some confusion on this message board. Currently, the folks pushing 40GbE are suggesting only a MMF (multimode fiber) and Cu (Copper) physical interface. There is a chance that a SMF (single mode fiber) 40GbE physical interface might be used, but it is unlikely to be embraced within the IEEE committee.

The 100GbE coterie is proposing Cu, MMF, & SMF physical interfaces. The 4x25G proposal has been only for SMF. Presentations have been made which suggest that the economics compare favorably -- you are welcome to use your own judgement in regards to these forward cost estimates. The 100GbE MMF proposal is likely to be a 10x10G interface both optically and electrically. It's an open question as to which MMF (40 vs. 100) rate is cheaper/bit.
Ryan Lawler 12/5/2012 | 3:05:18 PM
re: 40 GigE Could End Standards Spat Not at the IEEE meeting but still long to provide input into how you think the IEEE HSSG standards process should play out? Make your voice heard the democratic way -- through an LR poll on the subject:

http://www.lightreading.com/su...
Ryan Lawler 12/5/2012 | 3:05:18 PM
re: 40 GigE Could End Standards Spat Not at the IEEE meeting but still long to provide input into how you think the IEEE HSSG standards process should play out? Make your voice heard the democratic way -- through an LR poll on the subject:

http://www.lightreading.com/su...
tsat 12/5/2012 | 3:05:18 PM
re: 40 GigE Could End Standards Spat
Sure, it is doable. Pretty much anything is doable. But at what price?

-tsat
pavlovsdog 12/5/2012 | 3:05:18 PM
re: 40 GigE Could End Standards Spat provided standards can be established.

For example, a physical layer interface at 4x25GB using CWDM is readily doable today. All the circuitry behind it is also quite doable.

But having both 40GE and 100GE will scatter engineers and $, and jeopardize both.
Page 1 / 3   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE