02-28-2012 01:00 PM
Running an FCX stack of 2 switches, 220.127.116.11 firmware.
Quite a simple config, 16 virtual interfaces with ip policy route-map with 3 entries that matches traffic to one of 3 standard access-lists and routes to one of 3 gateways.
The switch is handling less than 1 GB of traffic throughput, however CPU usage is very high, averaging 75% and frequently spiking into the 90s, occasionally pegging at 99% for short periods.
doing a 'show processes cpu' does not tally, everything is idle except IP which sits between 20-25%.
The CPU load is real as we do occasionally lose packets and suffer degraded performance when the CPU hits 99%. PBR is the only feature we are using, all advanced routing features are off, no other access-lists applied, sflow disabled, very plain config otherwise.
I realize 18.104.22.168 is quite old, I will certainly be upgrading. Is there any thought to whether this will help or is PBR really sucha terrible memory hog? Also how can I get accurate CPU usage accounting since the process table is not showing me what is actually eating the CPU?
Thanks for any help!