CPU Maxing out to 98% over 5 minute intervals

[d.munro]

I have recently configured second serial interfaces on three MC3810's to try and achieve better throughput for the clients Video Conferencing. I initially configured ip policy routing to route the traffic that was destined for the VC bridge or gateway over the new PVC's that exist on the second 2Mb Frame Relay interface. I have also introduced a router at a remote site to this environment, this router had a new 2Mb FRA installed by the Comms provider and then another PVC on that link as well so that I could policy route the traffic from here also.

The problem that the three central routers don't appear to be doing any policy routing, but are sending a lot of traffic over these new PVC's and the remote router is perhaps 4 or 5 times a day maxing the CPU out to 98%. It is the IP Input process that appears to be causing this. I have been unable to determine whether it is the policy routing causing this or not. I have removed the route-map from the ethernet interface and it still happens. Do I have an unstable network? Any thoughts would be appreciated.

[millerv]

What switching method are you using ?

[d.munro]

I am using fast switching. I have considered using no ip route cache on the e0 interface but was unsure if this was necessary. The issue with the CPU on the remote router seems to have been abated, as my client has discovered that an ip address in the Netherlands was doing some form of DoS attack on them and have subsequently put a rule on their Netware BorderManager Firewall. Thankfully they stumbled across this. However this does not alleviate my issues with the traffic flow. I am also thinking that I may need to change the OPPF cost for these new PVC's as they have a higher bandwidth than the old ones that are intended for data, naturally. Any thoughts would be apprciated.