The Research Of Service Replacing Based Fault-tolerant Algorithm For Web Services

With the rapid development of Web Services fault-tolerant technology, how to create a low-cost and highly reliable fault-tolerant Web Services system has become the focus of attention. Traditional fault-tolerant technologies achieve fault tolerance mainly through creating redundant services, which lead to a large amount of redundant overhead. Web Services technology provides a unified specification which enables the client and server using different platforms and programming languages to co-operate. By using this feature, we can take equivalent services on internet to replace the traditional service redundancy to lower fault-tolerant costs. This article focuses on how to design and implement service replacement in the context of web sevices technology. The primary works are as follows:(1)A service replacing algorithm ARSR based on active replication is given to lower the cost that traditional fault-tolerant algorithm needed to establish redundant servers and to improve system efficiency. ARSR algorithm use the equivalent services on internet to create a service group. The members in the group deal with client request at the same time, and only the first returned response message will be sent to client. Design and implement a message coordinating algorithm to avoid nested call. It's strategy is as follows: Create a Hash table to ensure the uniqueness of the fastest member in service group. When a response message returned from a member, determine whether the member is the fastest member or not according to Hash table firstly. If the answer is positive, the response message will be sent to client and this member will be set as fastest member in Hash table. If the answer is negative, the message will be discarded to make sure that client will only receive the response of fastest member. Simulation results show that ARSR algorithm makes the operating speed of system depends on the fastest member. It improves the efficiency and reduces the cost of fault-tolerant system.(2)Because active replication algorithm needs to use equivalent services on internet as pre-overhead, a service replacing algorithm PRSR based on passive replication is given to lower the cost. PRSR algorithm also use the equivalent services on internet to create a service group. Only the primary server deal with client request when system runs. When fault occures on primary server, a backup server in service group is selected to continue responsing client request. Design and implement a fault recovery algorithm to maintain the consistency of system state. Because there are differences between primary server and backup server, such as system version and programming language, we can't cover the state of primary server before failure to backup server like traditional fault-tolerant algorithm did in fault recovery process. PRSR algorithm create a request list to store client request. When a backup server needs to be recovered, the requests in request list are replayed to it. The state of backup server will be consistent with primary server after dealing with all the requests in request list. Experiment and analysis results show that PRSR algorithm can reduce the cost of fault-tolerant system and the consumption of network resource effectively.