Web Server Caching vs Proxy Caching

There is a significant difference between document caching in primary and
proxy Web servers. In proxy servers, caching is done on a proxy server’s
disk (main memory is primarily used for meta-data and network buffers).
The set of requested documents is huge (potentially the whole World Wide
Web may need to be cached). Primary Web servers, on the contrary, need
to cache a relatively small set of local documents in main memory. Also,
primary Web servers have more skewed access patterns (popular documents
are requested much more often than unpopular ones). This allows for better
cache utilization. Direct application of the proxy caching policies to primary
Web servers may not result in the best performance.

In Web proxies, CPU overhead is not very important because almost all
cached documents are stored on disk and the time to maintain auxiliary cache
data structures is much smaller than the time to read a file from disk. In
primary Web servers, documents are cached in memory. Hence, the internal
complexity of the caching algorithm becomes an important issue.