This is a generic TCP/IP routing issue that significantly impacts the Internet Caches, as they open connections to huge numbers of hosts in the public internet. The problem happens if the cache is placed within the same IP subnet as the outbound internet router, but it is configured not to use it as a default gateway for some reason. To check if this is the cause, try to issue "sh ip route" command on the cache after some minutes of operation. If there are seen routes that were not configured, they are learned from ICMP redirect packets. Those routes are created for every host separately and
after a while the device spends all CPU time looking up the routing table before any packet can be sent out.
Solution:
If this happened by misconfiguration of default gateway,the solution is to apply correct configuration on the cache.
There might be a design requirement for this, if some Internet/Internal sites are reachable via different routers in the segment or a dynamic routing protocol is used to determine the routes. The best solution for this case is to change the network design so the cache appears on a separate subnet. In a case a L3-switch is used it can be done
by BVI reconfiguration. In some situations there might be appropriate to disable sending
the ICMP redirect packets by configuring "no ip redirects" at the interface of router used as default gateway by the cache. This generally causes less efficient usage on the local network, so it should be considered carefully.