[c-nsp] 1841 suitable for BGP?

[Mikael Abrahamsson]

On Fri, 1 Sep 2006, Everton da Silva Marques wrote:

> 2. the CEF path somehow executes at a higher priority
>   than processes like BGP and process-switching
>   (huh, is process-switching an actual process?);
>   hence CEF-switching performance would be less
>   affected by ordinary processes (compared to
>   process-swiching)

That is exactly the case.

CEF is interrupt-based forwarding, ie it'll interrupt any running process 
and do packet forwarding, when there are packets to forward in fast-path.

This works pretty much like it does on a linux box when it's servicing 
interrupt based tasks. Just because processes gets plenty of time to 
execute in the multitasking environment, doesn't mean it'll wait to 
service packets coming in on the network card (which probably has a very 
small buffer) or handle hd accesses or alike.

So BGP scanner might take 100% cpu for seconds without noticably affecting 
packets being forwarded through the device in question. If you see packet 
performance being affected by BGP scanner or other CPU process activity of 
low priority, open TAC case and get it resolved.

-- 
Mikael Abrahamsson    email: swmike at swm.pp.se