[c-nsp] 2621 with cef

[andrew2 at one.net]

cisco-nsp-bounces at puck.nether.net wrote:
> On Thu, 27 Oct 2005, Steve Snodgrass wrote:
> 
>> On Thu, Oct 27, 2005 at 03:43:12PM +0200, Horváth Szabolcs wrote:
>>> Is it possible that 2621's max. throughput is around 60-65 mbps?
>>> I tried various version from 12.0 to 12.2.
>>> 
>>> Pps is around 1.5-2k. Cpu is on 99%/98% (five seconds average).
>> 
>> You are lucky to get that much.  I have a 2621XM that I had to rate
>> limit to 40 Mbps so that it wouldn't drop its OSPF neighbors.  Cisco
>> specs the router at 1500 pps. 
>> 
>> 
> http://www.cisco.com/warp/public/765/tools/quickreference/routerperfor
>> mance.pdf
> 
> According to that chart, none of the above numbers make much sense.
> Cisco specs the 2620/2621 at 1500pps if you're process
> switching...but under normal conditions, it should be cef
> switching most of the traffic, in which case they say it's
> good for 20kpps.  Even cef switched, they say it's only good
> for 12.8Mbps...which in the real world probably means it'll
> have a hard time when you get close to 6Mbps.
> 
> Your 2621XM is spec'd at 15.36Mbps.  Unless the pdf is
> incorrect, I find it pretty hard to believe you have one
> pushing 40Mbps.

I routinely push ~60mbps through a pair of 2621's.  (In fact somewhere
in the cisco-nsp archives is a post where I try to find a way to squeeze
even more performance out of them :-D)  Keep in mind that's all traffic
in one direction, static routes only, and it's backup traffic so it's
probably a best-case type of scenario but I can vouch that it's
definitely possible to push >40Mpbs through 2621's.

Andrew