Application of game theory and fictitious play in data placement

Data allocation comprises data placement or migration into or between the data servers in a distributed data sharing system. In the process of data placement, static equations would mostly select only the best data servers. This would overload some of the data servers. An algorithm for optimizing file-placement in data grid environment (DGE) is presented. The algorithm adapts game theory and fictitious play to provide a mechanism for load balancing. The algorithm considers the overall placement behaviours and dependencies. A discreet-time based simulator has been developed to test proposed algorithm. Initial simulation result showed that game theory and fictitious play can improve the overall performance of the system.