[c-nsp] High interrupt load on 7507

Michael Loftis mloftis at wgops.com
Sat Jul 1 01:05:06 EDT 2006



--On June 21, 2006 4:12:46 PM -0400 David Coulson <david at davidcoulson.net> 
wrote:

> Rodney Dunn wrote:
>> Yes...faster CPU more pps. But it will be no where close to unicast
>> switching performance.
>
> What level of performance increase am I going to see by moving from RSP4
> to 4+, 8 or RSP16?
>
>> fact that it's not dCEF and the code isn't that efficient in the
>> switching path for bridging so performance is worse. To compensate
>> you can buy a faster processor.
>
> I might reshank the router so it's not running BVIs and see what
> happens. Throwing more hardware at the problem seems like a better
> solution :-)

Personally I'd do all my L2 Agg somewhere else....thats actually exactly 
what I do in most/all the places I have control over things.  Works lots 
better, way cheaper, unless you need weird interface types for some reason, 
in which case you probably should be re-thinking bridging anyway.  Routers 
are not good bridges.


More information about the cisco-nsp mailing list