[73807] in North American Network Operators' Group
Re: Internet speed report...
daemon@ATHENA.MIT.EDU (Lincoln Dale)
Fri Sep 3 19:22:23 2004
X-Envelope-From: ltd@interlink.com.au
X-Envelope-To: nanog@merit.edu
Date: Sat, 04 Sep 2004 09:19:42 +1000
To: Deepak Jain <deepak@ai.net>
From: Lincoln Dale <ltd@interlink.com.au>
Cc: nanog@merit.edu
In-Reply-To: <4138A113.207@ai.net>
Errors-To: owner-nanog-outgoing@merit.edu
At 02:51 AM 4/09/2004, Deepak Jain wrote:
>http://slashdot.org/article.pl?sid=04/09/03/0534206
>
>6.63 Gbps
>
>The article ended with hardware specs 'S2io's Xframe 10 GbE server
>adapter, Cisco 7600 Series Routers, Newisys 4300 servers using AMD Opteron
>processors, Itanium servers and the 64-bit version of Windows Server 2003.'"
>
>Is there a 10GE OSM for the 7600s?
there isn't any 10GE OSM but there are certainly 10GE modules for the
6500/7600 with deep buffers.
most recently, Guido Appenzeller, Isaac Keslassy and Nick McKeown have
written a paper that they presented at SIGCOMM 2004 about "sizing router
buffers" that is very informative and goes against the grain of the amount
of buffering required in routers/switches.
for folk who aren't aware of some of the hardware limitations faced today,
the paper provides a fairly good degree of detail on some of the technical
tradeoffs in router/switch design and some of the technical hurdles faced
by ever-increasing interface speeds today.
while Moore's law means that the processing speeds get faster and faster,
the same amount of innovation cannot be said for either speed-of-RAM or
chip packaging, which have fallen significantly behind both Moore's law and
speed-of-interface growth curves.
cheers,
lincoln.
NB. not speaking for my employer, this is an area of research that is of
personal interest to me. but of course, my employer spends a lot of time
looking at this.
