[nsp] "sh proc cpu" granularity

[Dennis Peng]

Charles Sprickman [spork at inch.com] wrote:
> Howdy,
> 
> After having gone access-list crazy last night, I see that one of the
> routers has started chewing more cpu.  This is expected, and I actually
> like the idea of the "infrastructure filtering", so I'd like to keep this
> after I upgrade.
> 
> This started me wondering though, how do I see how some of the features
> that don't have process entries (like ACLs) are affecting CPU and/or
> memory usage?  It seems that a whole mess of stuff gets lumped into "IP
> Input"...

Only traffic which is process switched will show up under the process
"IP Input". All traffic desinted to the router will be process switched.

Fast/CEF switched traffic through the router will be done at interrupt
level. If you look at "show proc cpu", the 5 second utilization is X/Y
and X represents the total amount of CPU time used and Y represents
the total amount of CPU time used at interrupt level. As your transit
traffic increases, the amount of time spent at interrupt level will
increase.

For more information, you can check out:

http://www.cisco.com/warp/customer/63/highcpu.html
http://www.cisco.com/warp/customer/63/showproc_cpu.html

Dennis

> Also, I once tried compiled access lists on this particular router and saw
> cpu dip down quite a bit.  Sadly, it also started randomly filtering
> things it shouldn't.  In general, is this a stable feature?  I'm stuck in
> the 12.2T train right now.
> 
> Thanks,
> 
> Charles
> 
> --
> Charles Sprickman
> spork at inch.com
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> http://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/