I have inserted a 3662 (256Mb 12.3(22) ) between our core and an internal pix intending to collect netflow statistics.
Fast0/0 801.q encaps
Fast0/0.86 --> core
Fast0/0.99 --> pix inside interface
Fast0/1 --> pix backend servers
Essentially all of the traffic (for now) consisting of low volume flows runs between the subinterfaces. I have enabled pbr on the core interface to push traffic sourced from the backend hosts to the backend of the pix.
CEF appears enabled by default on all interfaces.
Cpu utilization hovers about 20% and peaks about 50% (see attached proc cpu output)
1. Would enabling netflow likely push Cpu utilization over an acceptable level?
2. Would reconfiguring the interfaces as
Fast0/0 --> core
Fast0/1.99 --> pix inside interface
Fast0/1.11 --> pix backend servers
likely reduce cpu utilization?
Appreciate this and any other advice: thanks!
proc cpu threshhold