A Server Selection Method for Web-Based Multiserver Systems

Abstract

In our previous work, we proposed a dynamic data reallocation method for Web-based multiserver Multiplayer On-line Role-Playing Game (MORPG) systems. This method handles the data allocation as a combinational optimization problem. It is solved by using a tabu search that is a well-known meta-heuristic algorithm. Our system has some Web servers and an additional computation server. The computation server solves algorithm and decides the data allocation. However, each load of server is not completely balanced. This means that some Web server might have free computing resources. We aim to use these resources effectively. Therefore, our system handles a Web server that has free resources as a computation server. In this paper, we propose a selection method of a computation server in the Web servers. The server loads might be predicted because the data allocation is decided according to the last server loads and the user requests. Therefore, our system predicts the load of each server and selects a Web server. This selected server acts the computation server and decides the data allocation for the next term. After that, the computation server predicts the server load and selects the next computation server. We evaluates the performance of our system by measuring the latency values for game users and the server loads.