| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
PCIe gen 3 is expected to be 10 Gbps. The calculation would be 8 Gbps (unencoded data) * 8 (more typical lane count) * 75% (PCIe efficiency) = 48 Gbps. A 16 lane PCIe host bus would be able to handle about 96 Gbps which would be close to the maximum line rate of 100 GbE.
While the host bus may be able to handle that bandwidth, the CPU and memory will lag that bandwidth capability. Therefore, 40 GbE is probably sufficient for most servers over the 5-10 years.
Does this mean a task force should do a 40G MAC?
In my humble opinion, I don't believe that's necessary. What would seem to be a better effort is the development or enhancement of 802.3 link aggregation.
Four by 1G is very common today for adapters. It is my observation that most vendors would prefer a better link agg. Is that something worth considering as a new project?
Cheers,
Brad
-----Original Message-----
From: sanjeev mahalawat <sanjeevmahalawat@GMAIL.COM>
To: STDS-802-3-HSSG@listserv.ieee.org <STDS-802-3-HSSG@listserv.ieee.org>
Sent: Sun Apr 08 11:57:49 2007
Subject: Re: [HSSG] 40G MAC Rate Discussion
On 4/8/07, sanjeev mahalawat <sanjeevmahalawat@gmail.com> wrote:
PCIe 2.0 (released in January 2007) supports 5Gbps, x32. That comes out to be 32x5 = 160 Gbps (it does seem like a bottleneck for 100GE).
Meant to say "it does NOT seem like a bottleneck for 100GE".
Thanks,
Sanjeev
Thanks,
Sanjeev
On 4/8/07, Vandoorn, Schelto <schelto.vandoorn@intel.com <mailto:schelto.vandoorn@intel.com> > wrote:
Daniel,
I completely agree with you that we have to get moving on 100G ASAP, but we also need a platform to move the data or else history will repeat it's self as with 10G, where in the beginning, there was no platform to handle 10G and only with the introduction of PCIe, 10G started taking off.
To handle 100G we need PCIe GEN 3 and that's a long ways out. PCIe GEN 2 will be a nice fit for 40G and that's why we are proposing an additional 40G step on our way to 100G.
Schelto
________________________________
From: Daniel Cheok Kiang Kho [mailto: dkho@ALTERA.COM]
Sent: Sunday, April 08, 2007 3:40 AM
To: STDS-802-3-HSSG@listserv.ieee.org <mailto:STDS-802-3-HSSG@listserv.ieee.org>
Subject: Re: [HSSG] 40G MAC Rate Discussion
Jay,
Good point. The end users are already saying we're behind schedule for 100GbE, and are giving us the impression that they will buy 100GbE as soon as products become available. There will be a time frame of probably a few years before HSSG becomes a task force, for the task force to draft out the standard, and for the draft to go through IEEE approval. By then, I believe many applications will require 100GbE.
Regards,
Daniel
________________________________
From: Jay Moran [mailto: jm@ieee.org]
Sent: Sunday, April 08, 2007 7:05 AM
To: STDS-802-3-HSSG@listserv.ieee.org
Subject: Re: [HSSG] 40G MAC Rate Discussion
Thomas (and others),
Those end users who have posted to the mailing list are saying that we won't buy 40GE because we don't need it, and that 10GE (frankly 1GE) is all we'll need for our servers. I'm sure there are a few specific applications for greater than 10GE to a single server, but it is NOT for those of us running really large INTERNET data centers. I would still love to hear from end users requesting 40GE, since there are several of us requesting 100GE.
The only end users we've seen send information to the thread have been saying, stop wasting time fiddling with 40GE and get to where the real needs are... 100GE for switch->router and router->router links. We'll buy those as SOON as they can be fielded.
As for folks who have CFOs (or even CIO/CTOs) telling them what to buy, I strongly suggest you look for an employer where you have a say in technology adoption because there are lots of companies that do respect their internal experts and don't just buy what vendors put in front of them.
Jay
On 4/7/07, Thomas Dineen < tdineen@ix.netcom.com <mailto:tdineen@ix.netcom.com> > wrote:
The bottom line here is that if 40 Gig solutions become available in
volume over the next few years, which are perceived to provide good
customer value, they will be adopted by the customer! I giggle when
I hear statements like "We would never buy a 40 Gig solution"! Why?
Because I know that the CFO will override the desires of engineering
and buy the 40 Gig solution if it is cheaper. So there you have it 40 Gig
guys, just focus on schedule and value and you will succeed!
Thomas Dineen
--
Jay Moran
http://www.tp.org/jay