ICMP times to router

[AyrishGrl]

We have two 7204s for our border routers. The routers are connected to a Nortel Passport 8600 (layer 3 switch) and we have a PIX FW also off the switch. When I run a continous ping from our internal network to our dmz network on the router I am getting spikes in ping times. I have seen this when pinging the dmz interface from the PIX fw as well. Response times will go from 0-1ms up to 200ms for a series of 4-5 pings and then they will drop back to normal. Port stats for the fw and the router on the core switch are clean. I have worked with Cisco on this and they are telling me that this is due to ICMP having a lower priority in the Cisco processing engine. I realize this is the case, but I have never in my experience seen the ICMP low priority cause such regular spikes in ping times, especially when the router is far from taxed. These boxes run around 12% CPU. We have an OC3 coming into it and we don't use half that bandwidth. I havn't seen it spike over 8m yet. I just am not satisfied with the ICMP priority answer to this and it bothers me to see such regular spikes in response time. I am not sure what else to do to isolate any problems.

 

[AyrishGrl]

I just analyzed some of my ping data and it seems the spikes happen every 65-66 packets. This seems to regular to be a priority issue????

 

[vipergg]

Your only other option would be to see if you can see anything with a sniffer . Don't know what you are running for routing protocols , maybe this happens when doing a routing update , if you are running all statics then obviously that would not apply.

 

[AyrishGrl]

We are running iBGP between the two routers. One router is our primary and the other is its backup. Also have HSRP running. I am not great at BGP, but once it converges shouldn't the only updates be if a route changes? All other routing to our internal network is static. I am looking at a packet capture now, but so far havn't come up with anything.