WWW traffic reduction and load balancing through server-based caching

Most caching studies for large distributed information systems concentrate on client-based caching, which caches recently and frequently accessed information at the client (or at a proxy thereof) in anticipation of future accesses to that information (see the “Client-based caching research” sidebar). However, these techniques are myopic: they focus exclusively on caching at a particular client or set of clients, and hence are likely to have limited impact. The Object Caching Environments for Applications and Network Services (Oceans) group, part of Boston University’s Computer Science Department, has developed a more global solution that allows the replication of information on a supply-and-demand basis. This process is controlled by servers, which unquestionably have a better view of data-access patterns than clients have.