Cisco Support Community
Showing results for 
Search instead for 
Did you mean: 

Welcome to Cisco Support Community. We would love to have your feedback.

For an introduction to the new site, click here. If you'd prefer to explore, try our test area to get started. And see here for current known issues.

New Member

npe-g1, ip accounting, cpu problems


i am testing a npe-g1 in the moment and i have to use ip accounting outgoing-int on the box.

i use the 12.2(14)S IOS (also tested the 122-4.BW1) and when i increase

the ip accounting-threshold to 100000, the cpu goes to 100%.

i use a small DOS-tool to make some traffic, as we are an ISP and the router

will be DOSed, when he is in production. i generate 33kpps with spoofed


without the ip accounting, the cpu goes only th 16%, with ip accounting

nearly to 100.

CPU utilization for five seconds: 98%/98%; one minute: 98%; five minutes: 67%

PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process

it seams that if i make a clear ip accounting, all packets are process-switched and after a while, they hit the route-cache.

any hints how i can remain with an ip accounting-threshold of 100.000

and a normal cpu utlilization?

kind regards,



Re: npe-g1, ip accounting, cpu problems

You aren't process switching these, since all of your processor is taken up under the interrupt context (98/98). Check show alignment to make certain you aren't getting a ton of alignment errors or spurious accesses. If you aren't this box is just sucking up all its cpu in cef switching.


New Member

Re: npe-g1, ip accounting, cpu problems


here the output with the sh alignment. there are no alignment errors...

npe#sh proc cpu | e 0.00

CPU utilization for five seconds: 95%/95%; one minute: 32%; five minutes: 8%

PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process

npe#sh alignment

No alignment data has been recorded.

No spurious memory references have been recorded.

npe#sh int gi 0/2

30 second input rate 4964000 bits/sec, 10659 packets/sec

30 second output rate 4964000 bits/sec, 10658 packets/sec

but 100% cpu just for cef-switching? *ouch*