How to measure CPU utilisation

[PIKEYMIKEY]

Hi all, 

I've designed a couple of test topologies one running EIGRP the other OSPF and want to measure the difference in router CPU utilisation for each and compare. 

Does anyone have an recommendations on how to do this? 

Thanks 

[Philip D'Ath]

Unless you have a large number of routes (like thousands) the difference in CPU may be hard to measure.

You could try the command:

show process cpu history

[Masoud Pourshabanian]

You can also install a free or trial SNMP tool such as Solarwinds or Prtg to see the current usage or history through graphs

Masoud

[Philip D'Ath]

I don't think that would capture it.  Typically routing re-convergence happens on a sub-second basis unless you have thousands of routes.  SNMP polling would never pick it up.

[Julio E. Moisa]

Hi

Adding to Philip command:

show process cpu sorted, in order to see the utilization per process.

[Reza Sharifi]

Hi,

Unless there is a bug or something else is wrong with the IOS you are running, you are not going to see a significant CPU increase when running either protocols. You can always issue this command before and after enabling each protocol and if you don't see the protocols in this list then they are not consuming much bandwidth.

sh processes cpu sorted | ex 0.00

HTH

[Joseph W. Doherty]

As the others have already noted, normally (i.e. typical steady state - no topology changes) EIGRP or OSFP don't consume much CPU and normally the delta between them is probably not much.  Of the two, I would expect EIGRP to be somewhat less demanding than OSPF, because of OSPF's usage of SPF.  (The latter is also often mitigated by iSPF in later IOS implementations.)

Actual CPU consumption would be (much) impacted by routing protocol configuration and what's going on within your routing topology.

I believe both should have explicit processes that you could measure their on-going CPU usage.  Show process CPU should show those stats.

[Masoud Pourshabanian]

I have seen the CPU usage of 7600 to be between 40 and 100 percent but the problem was the network design. There were 700 routers in area 0 in an unstable network.