In today's bandwidth-hungry Internet, Active Cache's trading CPU for Internet traffic reduction is well worthwhile. Companies typically spend $6,000 to $10,000 per month for a T1 line to the Internet; a PC workstation that can function as a proxy costs less than $2,000 today.
Compared with traditional caching, Active Cache can increase cache hit ratios significantly. One source of the increase is the widespread use of cookies. Studies of HTTP traces from the AT&T WorldNet Internet Service Provider (1, 1) have revealed that 30% of all user requests carry cookies. Traditional proxies cannot cache replies to these requests because doing so would violate the semantics of the HTTP/1.0 protocol. As a result, they can only achieve object hit ratios around 35% and byte hit ratios around 30%. Active Cache proxies, on the other hand, can cache replies to requests carrying cookies with the help of server-supplied cache applets, and can achieve object hit ratios around 55% and byte hit ratios around 41%, according to the AT&T trace study.
Another source of the increase is the popularity of user-tailored information providers, such as ``my.yahoo.com.'' These information providers allow a user to choose among a collection of news and information items to be included in a customized Web page. The number of items is typically less than 1000, and each item is typically less than 1KB. Active cache proxies, at the help of applets from these Web sites, can cache the individual information items and provide the customized pages from the cache. Since each customized page is typically over 20KB, as long as the proxy has over 100 users, the cache applets will cut more than half of the traffic to these Web sites.
Though it is difficult to perform trace-driven studies of Active Cache since most proxy traces anonymize the URLs, existing studies do demonstrate the need and the potential benefits of Active Cache. Given the sharp contrast in the cost of computing versus the cost of wide-area communication, Active Cache will be an integral component of future Web caching technologies.