[c-nsp] Performance issue on link

Arie Vayner (avayner) avayner at cisco.com
Tue Apr 2 12:32:54 EDT 2013

+1. Sometimes the shaper does not match with the rate limiting the service provider uses on the link.
Reduce the shaper to substantially lower value (35M would be a good start, but maybe even 30M just to make sure).

If you see performance go up and work correctly at the shaped rate, start increasing the shaper until you hit the value that reduces performance.

You shaper rate has to be lower than the service policy the service provider is enforcing to avoid traffic drops on the circuit.


-----Original Message-----
From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of Mikael Abrahamsson
Sent: Monday, April 01, 2013 20:38
To: CiscoNSP List
Cc: cisco-nsp at puck.nether.net
Subject: Re: [c-nsp] Performance issue on link

On Tue, 2 Apr 2013, CiscoNSP List wrote:

> class-map match-any MATCH_ANY match any
> policy-map 40MB_SHAPE
> class MATCH_ANY
>  shape average 41943040

Try lowering this to 35 megabit/s to make sure you avoid the policer your provider has in place.

When you do the tcp testing, do a packet capture on the servers and look for packet loss. Wireshark has excellent TCP analysis tools that can show you exactly what's going on, if the problem is maxed out TCP windows or if it's caused by TCP saw-toothing due to packet loss.

Mikael Abrahamsson    email: swmike at swm.pp.se
cisco-nsp mailing list  cisco-nsp at puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

More information about the cisco-nsp mailing list