A cooperative model for wide area content delivery applications

Content delivery is a major task in wide area environments, such as the Web. Latency, the time elapses since the user sends the request until the server's response is accepted is a major concern in many applications. Therefore, minimizing latency is an obvious target of wide area environments and one of the more common solutions in practice is the use of client-side caching. Collaborative caching is used to further enhance content delivery, but unfortunately, it often fails to provide significant improvements. In this work, we explore the limitations of collaborative caching, analyze the existing literature and suggest a cooperative model for which cache content sharing show more promise. We propose a novel approach, based on the observation that clients can specify their tolerance towards content obsolescence using a simple-to-use method, and servers can supply content update patterns. The cache use a cost model to determine which of the following three alternatives is most promising: delivery of a local copy, delivery of a copy from a cooperating cache, or delivery of a fresh copy from the origin server. Our experiments reveal that using the proposed model, it becomes possible to meet client needs with reduced latency. We also show the benefit of cache cooperation in increasing hit ratios and thus reducing latency further. Specifically, we show that cache collaboration is in particular useful to users with high demands regarding both latency and consistency.