Research On Coordinator Election And Fault-Tolerance For MMOG

Currently, most of the commercial MMOGs are based on client/server computing model, which is lack of scalability, resulting in the communication bottleneck of game server. P2P based on DHT (Distribute Hash Table) technology offers great support for scalability of MMOG.. It has good abilities of scalability, reliability, uniformity of peer ID and self-organization. However, besides security aspect, P2P mode can’t offer a very support for MMOG because of the following problems of extending technology, the single-failure problem, the topology mismatching of P2P overlay, peer heterogeneity. They bring a big challenge to the election and fault-tolerance of coordinator. According to above problems, the main works in this paper been done as following:1. Propose a peer network distance prediction algorithm for P2P MMOGTogether with virtual region partition, the game world is divided into several measuring ranges. Coordinator calculates the network coordinates for player peers in one measuring range, and then clusters in those peers based on network distance, a newly joined peer can immediately calculate itself network coordinate according to measure RTT to cluster-center peer. A peer who has another peers’network coordinate can predict network distance, thus network traffic of measuring RTT is reduced, the algorithm can relax the topology mismatching problem of P2P overlay, it can also provide powerful help to coordinator election.2. Propose a coordinator election algorithm for P2P MMOGThe concept of peer utility is introduced. The factors that affect peer utility are given, including network topology information, capability, availability and reliability. The quantitative analysis of peer’availability is given. Network topology information is estimated by using peers’network distance. The calculation method of utility value is given, as well as reliability of peer. The coordinator can be elected according to the utility value. That can utilize computing and bandwidth resource better, and improve system performance, as well as increase the playability of game.3. Propose fault-tolerance mechanism of coordinator for P2P MMOG It is mainly discussed from the following parts: game data object and event, the hierarchical mode of peer, game replica recovery of coordinator. Game data objects are classified, so as to take the pointed measures to recover them after coordinator failure. The hierarchical mode of peer is proposed so as to the effective detection of coordinator failure. Adopting fully-replica redundancy, the player peers whose utility value is higher storage the game replica of coordinator. It can reduce continuous failure probability after coordinator failure. Coordinator makes event log and checkpoint for its temporary failure.The simulations show that the algorithm of peer network distance prediction has better precision under different quantity of landmark. And coordinator selection based on the algorithm of peer network distance prediction can reduce the average response value of coordinator in certain degree, and disperse system load effectively, improve game response performance. The fault-tolerance mechanism of coordinator can reduce the latency of replica recovery.