December 2001
Intermediate to advanced
400 pages
12h 2m
English
To return to our argument that the replacement policies are not the limiting factor of the progress in proxy caching, consider the proxy that showed the highest performance in a proxy performance evaluation conducted in January 2000 [Rousskov et al. 2000], capable of serving 2,400 requests per second. If the cache is provisioned properly, its average rate is at most a quarter of that, or 600 requests per second [Danzig 1998a]. At an average size of 10KBytes, this request rate translates into serving 48Mbps of data to clients. If we assume 40 percent of uncacheable data (see Section 6.3) that need not be stored and a 40 percent overall byte hit rate, the incoming data to be cached will arrive at the rate of ...