Performance comparison of active-sender and active-receiver policies for distributed caching

Abstract

We propose a distributed caching approach to offloading data access requests from overloaded data servers in a distributed system to nodes that are idle or less busy. Helping out the busy servers on data accesses, the idle or less busy nodes are called mutual servers. Frequently accessed data are cached in the main memory of mutual servers in addition to server and client local caches. We evaluate several data propagation strategies among data servers and mutual servers. Simulation results show that the Active-Sender/Passive-Receiver policy is the method of choice in most cases. Active-Sender policies are best able to exploit the main memory of other idle nodes in the expected normal condition where some nodes are overloaded and others are less loaded. All active policies perform far better than the policy without distributed caching.