Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
At 12:00pm west coast time, we held the second conference call to try to gain consensus on the speed objective for the High Speed Study Group.
(Apologizing up front for any spelling errors.)
Present on the first call:
Vipaul Bhatt
Ed Cornejo
Dan Ray
Paul Bottorff
Tom Dineen
Rich Taborek
Bruce Tolley
Mike Saltzman
Jonathan Thatcher
Walt Thirion
Present on the second call:
Ed Cornejo
Dan Ray
Piers Dawe
Ben Yu
Nader
Paul Bottorff
Roy Bynum
Tom Dineen
Rich Taborek
Jonathan Thatcher
Nan Chen
Bob Grow
Walt Thirion
Based on previous reflector discussion and the two conference calls, the group has pretty much settled into two choices for the speed objective, 10.000 Gb/s and 9.58464 Gb/s. A couple of straw polls indicate support for these two speeds is pretty evenly divided. Moreover, it isn't clear that either group will change within the next week or two.
Having said that, it is clear there is consensus among the HSSG-Speed ad hoc for these two speeds. These speeds are measured at the 10 gig MAC and, therefore, are independent of line codes, baud rates, line rates, etc.
The 9.58 proponents feel strongly that
(1) this speed will enable an easier, cheaper connection to the WAN infrastructure without unduly penalizing performance.
(2) a significant amount of research has already been done with components for this speed and we should leverage that research
(3) 9.58464 is a magic number in that it is the data rate for the payload in the SONET system
(4) The decision to use a 10.000000 Gbps data rate will limit 10 GigE to campus LAN applications. The interfaces required for adaption to the installed base of lit wide area networks will make Ethernet more expensive, more complex, and less scalable than competing solutions which are taylored to wide area networks. Ethernet at 10.000000 Gbps will unable to compete in the wide area.
The 10.00 camp is concerned that
(1) adopting the 9.58 speed will allow feature creep (with associated cost and complexity) and is a move towards adopting the perceived higher cost of the SONET physical layer.
(2) a speed less than 10.000 will not allow appropriate aggregating of lower speed links
(3) because ethernet has historically moved by factors of 10, not doing so this time opens a Pandora's box of issues on marketing, product acceptance, etc.
So, with the help of Bob Grow and Paul Bottorff, I'll put together a short presentation summarizing the major issues and ask for a time slot at next week's plenary. Then, I'll draft a motion that essentially says we will focus on these two speeds as the primary objective with the hope that further technical research, line code decisions, etc. will help resolve which speed survives and gets standardized.
Walter Thirion
Level One Communications
512-407-2110