[c-nsp] Sampled netflow on 6500/7600

[Richard A Steenbergen]

On Sun, Jul 02, 2006 at 11:08:41PM +0100, Ian Dickinson wrote:
> > Lower the normal aging timer, turn on fast aging.
> 
> Ok so I use:
> 
> mls flow ip destination
> mls aging fast time 4 threshold 2
> mls aging long 900
> mls aging normal 32
> 
> Is this sensible?  This just to try and make it calm down...
> What's the expected hit on CPU?

I'm running:

mls flow ip destination
mls aging fast time 10 threshold 16
mls aging long 240
mls aging normal 45

And on a 3BXL box with no DFCs and enough traffic to take tcam utilization 
to 80% (only a few gigs), I'm seeing:

rotuer#remote command switch show proc cpu sort 1min

CPU utilization for five seconds: 24%/8%; one minute: 32%; five minutes: 32%
 PID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process 
 242  1814246696 120720632      15028 11.75% 13.85% 13.96%   0 NDE - IPV4       

Not seeing any cpu time used on CFCs obviously. If you have DFCs, does it 
distribute the NDE cpu load onto them or is it still handled by the sup 
cpu?

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)