中文摘要：

基于对等网络（peer-to-peer）的大型多人在线游戏（massive multi-player online game,MMOG）是网络游戏研究的热点之一.其关键任务是建立有效的兴趣管理机制.大量研究采用兴趣域内单协作者管理的模式来组织节点.随着兴趣域内玩家数量的不断增加,协作者容易达到性能瓶颈.提出一种负载均衡的结构化多代理节点模型.该模型依靠节点加入算法和邻居发现算法来建立节点间的联系.同时通过计算资源消耗指数和节点性能指数,将域内资源的处理权均衡地分配给所有节点.由于资源状态在任意时刻都由唯一的节点进行更新,保证了所有资源状态的一致性.实验表明,该模型在多方面具有良好的性能优势.

英文摘要：

The massive multi-player online game based on peer-to-peer overlay is a research hotspot of network game. To establish an effective mechanism of interest management, most of the approaches organize nodes by using single collaborator to manage the area of interest. As the number of players and resources increases in an area, the collaborator would catch the performance bottleneck. This paper proposes a structured multi-agent model which divides the world into static areas of interest and introduces a manager to organize nodes in each area. Other nodes acting as agents are responsible for processing events. To make interactions between nodes and resources in the same area and neighbor areas, SMA （structured multi-agent） implements the node joining algorithm and neighbor discovery algorithm to establish connections between all nodes. Meanwhile, it calculates the cost indices of resources and performance indices of nodes, so that each node can hold several resources whose sum of cost indices is quite with its performance indices. By this way, resources are assigned evenly to all nodes in an area. Since the state of resource can only be updated by one node at any time, the consistency of all states are greatly guaranteed. Experiment shows that this model has fine advantages in many parts.