Dynamics of a Utility Based Distributed Video Proxy-Cache

We present a distributed video proxy-cache system that is able to adapt the number of running nodes depending on the client request patterns and available storage and computational resources. Cache related operations are performed with regard to the utility values of the stored objects. We propose the use of a new utility formula which considers a number of object characteristics. This formula is fine tuned by using a genetic algorithm for determining the best approach to weight those characteristics so that the yielded byte hit rate of the system is maximal. Also, a new approach that uses the standard deviation of the utility values for the cached objects, to increase or decrease the number of active nodes within the proposed system, is presented.