The Research Of Load Balancing In P2P-based MMOG

Many algorithms have been proposed to solve the load balancing problem in P2P DHT-based network. However, these algorithms ignore the heterogeneity and chur of network nodes, or affect the cost and speed of load balancing when the load is transferred between nodes without considering their close relationship, or use a centralized coordinator-based dynamic load balancing algorithms, but this algorithm may issue a single point of failure. For the above issues, the massively multiplayer online games system require a relatively high, So we do the following work to slove the above shortcomings of this article.1) The current load balancing algorithm not considering the relationship between neighboring nodes affect the load balancing cost and speed. So a load balancing algorithm considering the relationship between neighboring nodes is proposed. The algorithm base on virtual server and improve the transfer strategy to many-to-many mode. When a node is overloaded, we use the greedy algorithm to find the nearest node which is able to share its load in oder to make the transfer of load between nodes to reduce consumption.2) Ralizing the load balancing algorithm considering the relationship between neighboring nodes, We design a global distribution of K-ary free based on DHT, call the LIB module to collect the system information along the K-tree from the leaf node to the root, call NTS module to identify overloaded nodes, call VSA module to generate the deploy information,and call VST modules based on VSA information to transfer load.3) We design a simulation program to simulate the proposed load balancing algorithm. In the program,the structured overlay network is created by the Peersim. We ralize the KTnode class inherited from the node interface to simulate the proposed load balancing algorithm and prepare the configuration file, statistical analysis class and protocol class to assiste the emulation program.The improved algorithm and Proportion algorithms not considering the relationship between neighboring nodes are running in the topologyl and topology2 respectly. Experimental results show that the proposed method of load balancing is largely to reduce costs and able to speed up the load balancing speed.