Fault-tolerant Redundant Middleware Services In A Distributed System Design And Realization

With the widespread application of distributed technology, the environment has become extremely complex especially in the key areas, heterogeneous network is also severer, system need stronger reliability and availability, and fault-tolerence is the best way to ensure the performence. It is heavy and hard to develop fault-tolerence on the distributed application directly because of the complicated application logic and fault-tolerant logic. in order to reduce the burden of application developer, it is easier and smarter to provide fault-tolerence in the middleware level.Following the present fault-tolerent standard and learning some fault-tolerent ideas in other field, this paper proposed a fault-tolerent module and framework. Designing and proving the repliation manager and fault detection is the core mission in this paper, the repliation manager administrated replications by object group. Generic factory creates and deletes the objcet according to the request from replication manager. The system keeps the state synchronization between those replications on the basis of replication style. It proposed a pull/push mixed pattern in the fault detction module and could carry out the process-level, object-level and host-level fault detection using hierarchical detection.Test results show that the request average latency will be not bad in the 10/100M Ethernet local area network, if the message length is not more than 3k and the number of replications is less than 4, the fault detection coulddetect the fault in time and accurately.