Research On The Problem Of Web Service Failure

Service-oriented-architecture is a new distributed computing technology. Along with the rapid development of web service, there are increasing number of web services on the network, which provide similar functions for users. Therefore, Web Service selection based on QoS(Quality of service) has become a critical subject, but firstly, we should promise that the QoS evaluation result is real, which in fact, is difficult. Due to the flexibility of web service, it cannot work normally, although intuitively the approach has ensured the best performance choice. Nowadays web service failure problem has become a serious factor influencing the web service quality.By analyzing web service technology and QoS relevant knowledge, it is known that the web service release mechanism lacks of information warranty, and environments vary among different groups of network, and QoS changes dynamically. As a result, the selected services do not have a high availability. However, in this paper, some effective measures are taken to solve the problems as follows.First of all, taking the node failure (such as server and hardware breakdown) into consideration, we send a message to the web service regularly in order to verify the service available or not. More over, we present a web service availability prediction model based on Markov, which is used to deal with the above data. In addition, as far as the hot bottleneck concerned, the service might failure when the number of end-user exceeds the server capacity. Therefore, we present a queuing model, which could predict the probability of end-user overflow. On the whole, combining with the above mentioned predict result, we can draw the conclusion that which service has the higher availability.The experiment shows that the presented prediction model is efficient to achieve correct prediction, and it is an effective way to avoid service failure by predicting the availability before usage. On one hand, it helps requesters choose the service with higher availability. And on the other hand, it prompts providers enhance their performance and get more benefit.