>From: Avi Freedman <avi@freedman.net>
>To: rbuchals@hotmail.com (Ron Buchalski)
>CC: simonl@rd.bbc.co.uk, cisco-nsp@puck.nether.net
>Subject: Re: [nsp] RSP/VIP performance question
>Date: Fri, 10 Sep 1999 11:13:33 -0400 (EDT)
>
>
> > VIP2-40 can switch up to 65kpps (internet average packet size of 280
>bytes),
> > and has a total throughput of between 300 and 360 Mbps. I prefer to use
>it
> > conservatively, and engineer for 300Mbps. Any bursts above that are
>brief
> > and can be accommodated. Don't engineer for full-time 360Mbps or you'll
> > have a busy pager!
>
>That seems a bit high; 200Mbps-250Mbps seems more reasonable for a 2/40.
I'm not making up the numbers, Avi. Those are the numbers that are used in
the configuration guidelines for the VIP2-40. Of course, it's dependent on
packet size, since the 65kpps limit can be reached at a much lower
throughput level if the packets are small.
(64 Bytes) * (65 kpkt/sec) * (8 bits/Byte) = 33.28 Mb/s
(280 Bytes) * (65 kpkt/sec) * (8 bits/Byte) = 145.6 Mb/s
(576 Bytes) * (65 kpkt/sec) * (8 bits/Byte) = 299.52 Mb/s
(1500 Bytes) * (65 kpkt/sec) * (8 bits/Byte) = 780 Mb/s
Or, looking at it another way:
{[(360 Mb/sec) / (8 bits/Byte)] / [65 kpkt/sec]} = 692 Bytes/pkt
So, for packet sizes below 692 bytes the throughput will be limited by the
packet switching capability of the VIP2-40. For packet sizes above 692
bytes the throughput will be limited by the bandwidth of the VIP2-40.
These are raw performance numbers. Turning on software features which
require CPU resources from the VIP2 will reduce the CPU cycles available for
switching packets.
-rb
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:05 EDT