[c-nsp] c3560 L3 throughput with BGP

Łukasz Bromirski lukasz at bromirski.net
Tue Oct 30 16:06:54 EDT 2007

Adrian Minta wrote:

> On 3550 average CPU load was below 1%. On 3560 the load was always above 
> 5% with spikes up to 60%.
> After some googling I change the sdm template. The load went down, but 
> above 5%. I put extensive ACL's on all the routed port and finally CPU 
> stabilized to around 5%. The load is caused by "IP Input" and "HL3U bkgrd".

IP Input as always is responsible for processing traffic that
hardware/CEF isn't capable of processing.

'HL3U bkgrd' is actually managing a bunch of things (hardware ARP
throttling, proxy ARP, ICMP redirect, TTL ICMP error generation, etc).

So, I really suspect You did change switches but in meantime Your
network also changed traffic characteristics and that's why You see
the change in load.

"I am a part of everything that I have  |               Łukasz Bromirski
  read." -- Theodore Roosevelt           |    http://lukasz.bromirski.net

More information about the cisco-nsp mailing list