Home Based Cooperative Web Caching

Web caches in a cooperative cache array must have a method to determine if another cache in the array has an object corresponding to a specific URL. This is achieved by a broadcast query in Internet Cache Protocol (ICP) based cache arrays. An alternative approach is taken in Cache Array Routing Protocol (CARP), where a particular URL can only be cached in a designated cache. Both approaches have their disadvantages. A broadcast query is expensive and limits the scalability of a cache array protocol, while allowing an object to be stored only in a specific cache prevents the web clients from benefiting from locality. This paper proposes a new cooperative caching protocol called HOME which avoids both these drawbacks. Caches in HOME store all objects but give preference to home objects. Every web object is a home object for one of the caches in the array, and the identity of the home cache is based on URL hashing. When looking for an object in the array, only the home cache is checked. We demonstrate with trace driven simulation that HOME locates objects in the array effectively and achieves performance on par with cooperative ICP arrays, but without generating the large number of inter-cache queries associated with ICP query broadcasts.