Research And Realization Of Distributed System Fault-tolerant Middleware

The fault-tolerant middleware separates the fault-tolerant logical andthe application logic by providing fault-tolerant in the middleware levelfor the distributed system. At present, the domestic and foreign researchof the fault-tolerant middleware technology uses the followingfault-tolerant models, the integration model, the interception model andthe service model, which have their respective flaws. The existingfault-tolerant technology and the fault-tolerant middleware has thoroughstudied, and has proposed an improved framework, the pluggablefault-tolerant middleware PFTM, which is based on the pluggableprotocol framework PPF of CORBA.Firstly, the fault-tolerant CORBA standard and the pluggableprotocol framework PPF are analysed, and also the characteristic of eachfault-tolerant model is studied, such as the portability, the interoperabilityand the performance, and then an improved fault-tolerant middlewareframework PFTM is proposed. Secondly, according to the function of thefault-tolerant middleware, the server side and the client side fault-tolerantplug-in's structure and function of PFTM have designed. Combing thecharacter of the active repliacation and the passive replication inimplement object redundancy, the semi-active replication is proposed.The semi-active repliacation has overcome the question of the highexpenses in the network that brought by the active repliacation's massivemessages and the question of long failure recovery time in the passivereplication. Also the solution of the flow control question in the activerepliacation is proposed. Thirdly, the implement of the key fault-tolerantmechanism in PFTM is elaborated, including the replication management,fault detecting, the log and recovery and so on. Finally, the prototypesystem has tested and its performance has analyzed. A pluggable fault-tolerant middleware PFTM is proposed, which hasovercome the deficiency of the integration model's portability and theservice model's performance, and has solved the difficulty inmaintenance replicas' consistent with the interception model and servicemodel. Combing the service model and the interception model, using PPFto integrate the fault-tolerant mechanism into ORB, by which PFTMrealizes the strong replica consistent. The development of thefault-tolerant applications based on the PFTM does not need to reviseORB, only need to apply the extremely few revisions to the application.PFTM can provide transparently fault-tolerant for the application, and canprovide the interoperability between the ORBs which support PPF. PFTMcart use the semi-active replication to implement fault-tolerant, so itsfault-tolerant performance is close to active replication, better than thepassive replication.