A FUZZY BASED ADAPTIVE PREDICTION FRAMEWORK FOR ENHANCING THE AVAILABILITY OF WEB SERVICES FORINTERNET APPLICATIONS

The internet based applications are primarily supported by web services( ws) . The number of requests received on ws are varying dynamically from time to time which leads to slow down in response time during peak load periods. In order to overcome this problem, the availability of web service is increased by replicating the web services over physically distributed servers. In this paper we have proposed an adaptive prediction framework which uses Poisson and exponential distribution models to meet the Quality of Service(Qos) attributes (availability, response time) of web services under the high number of service request via service replication. The Poisson and exponential models are used to find the probability of request arrival rates, response time of reque sted services.It alsouses FL for efficient and Resource Control Algorithm (RCA) for decision making to determine the replication requirement. Simulated test environments have been created to evaluate the performance of our framework and test results are co mpared with existing models. The results confirms that our proposed framework maintains the response time of ws within expected SLA (Service Level Agreement) response time even during peak load and it significantly improves theavailability of ws with mini mal intervention of system administrator.