[c-nsp] 7200vxr-npe400/512mb - how much BGP?

[Ibrahim Abo Zaid]

I agree with Justin , currently it seems you don't have any memory problem
but you need to worry about box CPU especially BGP isn't the only active
process here and you need to monitor processor utilization closely and if
you faced sporadic peaks you can use show process cpu sorted command to
catch up the process eating the resource and isolate the peak trigger , is
it BGP scanner or IP Input process etc , ..

and finally in such cases there are some processes appears as the
*reason*behind high CPU but actually those ara
*results* of other causes so these problems needs accurate investigation and
always check IOS caveats as sometimes processing problems yields of coding
caveats


On 4/13/08, Justin M. Streiner <streiner at cluebyfour.org> wrote:
>
> On Mon, 14 Apr 2008, Skeeve Stevens wrote:
>
> > Just how much BGP should a 7200vxr-NPE400 with 512MB of RAM be able to
> > handle.
> >
> > The router currently says "Total: 466497056, Used: 200153224, Free:
> > 266343832"
> >
> > When should I start worrying about how big the tables are growing and so
> on?
>
> 512 MB is the minimum I'd consider using for a router that will be
> carrying full BGP feeds, but in this case, the limiting factor might not
> be memory availability, but rather the CPU, since everything in the 7200
> series is done in software.  Do you notice your CPU usage spiking
> periodically (around once a minute), and is a large chunk of the CPU tied
> up un the BGP Scanner process?
>
> If you have a tool for graphing and trending stuff like that over time
> (MRTG, Cricket, many others), you may want to set up something to monitor
> that CPU utilization, paying attnetion to both the 5 second and 5 minute
> CPU utilization values in the MIBs.  The 5 second value will help you
> catch transient spikes that get washed out of the 5-minute average
> values.  The output ends up more closely resembling the output of "show
> proc cpu hist".  When the utilization starts regularly getting close to
> 100%, it's time to think about an upgrade.
>
> I wouldn't worry so much about one or two errant spikes, but when things
> regularly get that high, it could manifest itself in the form of increased
> latency in getting traffic through the box, or if things get bad enough,
> the router starts missing BGP update messages or similar messages for your
> IGP, and sessions/adjacencies can start dropping, which only makes the CPU
> thrashing problem worse.
>
> jms
> _______________________________________________
> 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/
>