Your correct for traffic switched by the PFC3x it is 30Mpps for IPv4, and 20Mpps for MPLS. The reason MPLS is lower is for load balancing, you go and look at the underlying packet to see if it is IPv4 and use the src+dst IP address to compute the hash for the next hop, which results it looking deeper in the packet. To be precise
PFC3x IP to Label 30Mpps (push up to 3 labels) Label to Label 20Mpps Label to IP 20Mpps (pop 1 label) DFC3x IP to Label 48Mpps (push up to3 labels) Label to Label 48Mpps Label to IP 48Mpps (pop 1 label) DFC3x has wider path to do the lookup compared to PFC3x so the performance impact is not present. Ian On 7/1/10 5:20 PM, Richard A Steenbergen wrote: > On Thu, Jul 01, 2010 at 09:43:45PM +0200, Gert Doering wrote: >> Hi, >> >> On Thu, Jul 01, 2010 at 09:17:34PM +0200, Mikael Abrahamsson wrote: >>> What about 7600 and re-circulation? I don't remember exactly what is what, >>> but wouldn't there be cases where you'd get 15Mpps with regular CEF and >>> 30Mpps with label switching (LSR) ? >> I think it was the other way round, 30 Mpps for regular CEF and 15 Mpps >> for recirculated MPLS-in-pop-last-label-IP-out. > IIRC it's 30Mpps for regular IPv4 CEF vs 20Mpps when doing MPLS lookups. > Cut that in half if you have to recirculate (two lookups), etc. > _______________________________________________ cisco-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
