[c-nsp] 1841 suitable for BGP?

[Everton da Silva Marques]

On Fri, Sep 01, 2006 at 10:55:08PM +0200, Gert Doering wrote:
> 
> > > I'm looking at a router to handle 2 x 10Mb/s (ethernet)
> > > feeds and take a full BGP routing table from both ISPs.
> > > Will an 1841 with 384Mb suffice for this, or do I need
> > > to look at higher spec models?
> > 
> > Technically yes, it will handle it BUT (thats a big but)
> > the 1841 is not a distributed router so the main CPU has
> > to do everything.  The BGP updates once per minute will
> > crush the router and cause all packets to stop.  
> 
> What you describe is true for process switching, but if
> you *are* doing process switching, you have much worse
> problems than the BGP scanner.

I used to assume that the 1841 router was deprived of
any ASIC-assisted forwarding mechanism (think TCAM-like),
and thus packet switching would be fully performed by
software in the main CPU; the performance distinction
between process switching and CEF switching would lie
entirely in the software path triggered by the incoming
packet. That is, only the low priority of processes
like BGP scanner would prevent those processes from
routinely disrupting the CEF switching performance.

I suppose you are saying that either:
1. 1841 does have some kind of specialized hardware for
   packet forwarding, presumably programmed from CEF FIB
--OR--
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)
?

Please clarify.

Cheers,
Everton