[c-nsp] Routing table

Stephen J. Wilcox steve at telecomplete.co.uk
Wed Jul 27 12:32:55 EDT 2005


On Wed, 27 Jul 2005, Jean-Christophe Varaillon wrote:

> > > The CPU is flirting with the 35% at peack hours.
> > >
> > You never mentioned what platform this is, but for older platforms BGP
> > Scanner itself can easily add 10-15% (on 5 minute averages) of
> > background noise.
> 
> 7206VXR (NSE-1), handling 40Mbps of traffic.
> 
> > What's your utilization break down of interrupt vs. process cpu?
> 
> I am not sure how to check this : /

#sh proc cpu | e 0.00
CPU utilization for five seconds: 4%/3%; one minute: 11%; five minutes: 11%
 PID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process 
  17    78130256 502561569        155  0.08%  0.08%  0.12%   0 ARP Input        
 114   124632820 226266944        550  0.24%  0.09%  0.08%   0 BGP Router       
 115    39529212  82668310        478  0.24%  0.04%  0.01%   0 BGP I/O          

the '4%/3%' means a total of 4% cpu with 3% from interrupts

doing '| e 0.00' means i get to see a short list of the major cpu hogs


> > Are the 'internal routers' participating in BGP or are the edges injecting
> > a default into the IGP?
> 
> There are three ISPs, two ASBRs and two internal routers.
> The internal routers are route-reflector clients receiving the most usefull
> routes only (#2000 prefixes).
> The internal routers are defaulting to a network, learnt via BGP as well.

i'd be tempted to run all 4 routers as ibgp peers and drop the route reflection, 
this may reduce the number of bgp updates thro the system as the reflection 
means that all updates are reflected


if you do 'sh cef int' just confirm that all your interfaces are listed.. 

Steve



More information about the cisco-nsp mailing list