[45924] in North American Network Operators' Group
Re: Line Rate vs CPU utilization info requested
daemon@ATHENA.MIT.EDU (Stephen Sprunk)
Thu Feb 28 11:32:46 2002
Message-ID: <00ea01c1c075$61710cf0$e1876540@ssprunkpc>
From: "Stephen Sprunk" <ssprunk@cisco.com>
To: <nanog@merit.edu>, "Donn Lasher" <dlasher@clearskynet.net>
Date: Thu, 28 Feb 2002 10:28:08 -0600
MIME-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Errors-To: owner-nanog-outgoing@merit.edu
Thus spake "Donn Lasher" <dlasher@clearskynet.net>
> Does anyone have a table / web site / etc that talks about realized IP
> throughput on various routers (3600 / 7100 / 7200, 5500, 6500, etc)
> versus cpu load? IE even though you can put 4 x 100M ethernet cards
> in a 3640, IMHO, you'd never see 400/800M aggregate throughput
> because the CPU would be on its' knees begging long before that point.
The real-world limitation for a 3640 is 70kpps (no acls, fast sw, 100% cpu).
At 45Mb/s, a 3640 should be just warming up.
> I'm looking at a project that calls for a DS3, and the suggested router
> is a 3640, and I'm trying to prove / document that's not a wise idea.
A 3640 should work fine for that application, provided you don't have any
other major requirements like BGP, long ACLs, etc.
S
