[Venkatesh]

How about using the Squid cache in http-accelerator mode on Mike's server? That way, you can leave the current configuration in place and take advantage of both extra bandwidth and reduce processing when people are dealing with both static elements and threads which haven't been touched in a while. And the configuration of that is a lot simpler than moving servers.

To balance load between any number of volunteer caches and the primary server, you could us one of many load balancers (pen, plb, etc.) or round-robin DNS. Then each cache server would run squid and set itself to be caching chiefdelphi.com. When the squids can't serve requests, they will pass them back the the current server.

That might help some with performance.