On (2008-02-08 13:04 +0100), vince anton wrote: > PFC bus utilisation from 'sh catalyst6000 traffic-meter' on my box shows me > 8%, presumably 8% utilisation of my 30Mpps. But im guessing the 30Mpps is > based on lab conditions with 64 byte packets right ? > > So what a fair measure for real world PFC utilisation with internet transit > traffic, (which is of course not exclusively 64 byte), which can be used to > indicate thats its time to look at DFCs ?
The pps limit of EARL7 is packet size agnostic, only headers are transmitted. And actually it's not EARL7 that craps out at ~30Mpps, it's the bus, the bus is fully saturated with pushing packets for EARL7 to do it's lookup. I think it's like 32btes per cycle, IP is 2 cycles and MPLS is 3 cycles (since it needs IP if you do ECMP). Bus is 62.5MHz, I think. So from this we'd arrive at: (62.5*1000000)/2/1000000 = 31.25Mpps IP (62.5*1000000)/3/1000000 = 20.83Mpps MPLS This of course assumes that bus will only have headers for lookup, that is, you don't have classic LC's that use bus for transferring data. > > > > Thanks, > > anton > _______________________________________________ > cisco-nsp mailing list [email protected] > https://puck.nether.net/mailman/listinfo/cisco-nsp > archive at http://puck.nether.net/pipermail/cisco-nsp/ -- ++ytti _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
