Implementing Domain Caching

Web servers providing live coverage to events are hard to design due to a high degree of variation in the load, variation in geographical distribution of users and the frequency with which content is updated. In this paper, we describe an implementation of domain caching, a technique that dynamically distributes the load on the server by instructing selected clients (domain caches) to cache server content and service requests from nearby clients. We built a prototype domain cache that can be used by clients willing to act as domain caches. We evaluated domain caching using this prototype and found that accessing content from a nearby domain cache can be up to an order of magnitude faster than accessing the same content from the origin server. We also show that our prototype is able to handle the load expected to be handled by most domain caches, with only a modest increase in the execution time of other applications. Ac-cessing the prototype from several clients located geographically near the domain cache suggests that geographical proximity could be used to identify nearby clients. Finally, we showed that by incorporating a soft mapping from the server to the domain cache, clients can handle domain cache failures. We addressed security implications of domain caching by discussing how clients can use digital signatures provided by servers to authenticate content.