[48057] in North American Network Operators' Group
RE: Cisco 7200 VXR with NPE-400 (was RE: The market must be comin
daemon@ATHENA.MIT.EDU (Ralph Doncaster)
Wed May 22 10:13:52 2002
Date: Wed, 22 May 2002 10:14:46 -0400 (EDT)
From: Ralph Doncaster <ralph@istop.com>
To: "Heath_Dieckert@Dell.com" <Heath_Dieckert@Dell.com>
Cc: "nanog@merit.edu" <nanog@merit.edu>
In-Reply-To: <E8B7DBCA000B1247B8524630773472C6B90C74@AUSXMRR201VS1.aus.amer.dell.com>
Message-ID: <Pine.LNX.4.21.0205221010390.3024-100000@cpu1693.adsl.bellglobal.com>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Errors-To: owner-nanog-outgoing@merit.edu
> Based on our testing it looks like it all has to do with packet size. With
> small packets the throughput is very low. With what Cisco calls an
> "internet mix" of packet sizes throughput is much better. When doing max
> MTU packets, the throughput is of course the best.
The other thing I've found about traffic type is how sensitive
netflow is. I was running it for a while, then I got a co-lo customer
that had a lot of UDP traffic with small packet sizes and rarely more than
a few packets between the same src/dest ip/port (much like DNS
queries). It was enough to flatline the box and cause it to crash.
-Ralph
