Selecting the Optimal Web Service Composition Based on QoS

Abstract

In recent days, the fast spread of web services in our businesses and day-to-day lives has made QoS a very important aspect for both the service provider and the consumers. The consumers prefer to deal with the best web service available on the internet. The main problem is how the consumer obtains a high comprehensive quality composite service when there are a large number of web services available, so it is not easy to use an optimal execution path of web services; the choice of the optimal path depends on the QoS for every atomic service. The values of QoS attributes which are published by service providers may be not trusted or inaccurate; some of the developers typically measure a QoS in terms of response time, throughput, and availability. Our contribution is to study the in uence of the reputation factor in the process of selecting the optimal path in the absence of one of four factors (Availability, Reliability, Response Time, and Price) and the possibility of covering for them. We have used the reputation factor when calculating the QoS by using arti cial bee colony algorithm for selecting the optimal web service composition. Then we analyzed the impact of reputation on the process of selecting web service composition in terms of the QoS and accuracy of the solution. Also, we studied the impact of the reputation factor in the case of the absence of one of the four factors through three experiments and a set of comparisons. The result was that the reputation factor could cover for factors such as availability, Response Time, and technical support. We used multiple linear regression and polynomial regression to show the prediction of the reputation factor using the four other factors. The result has a higher con dence when we used multiple polynomial regression where the Residual Sum of Squares (RSS) was less than the multiple linear regression. In addition, we analyzed the association between reputation and the four other factors using ANOVA test; the result indicates that there is a signi cant association between reputation and (availability, response time, and price), but it is not signi cant association with the reliability