Research On QoS-Aware Web Services Composition Based On Genetic Algorithm

With the rapid development of network and web services technologies, those traditional application intergration models which perform tasks by relying on experts to compose web applications acoording to fixed plans are no long fit to current web enviorment. Therefore, researchers are now focusing on how to compose these current useful web services automatically in order to get more powerful and value-added service processes which can satisfy users' requirements better.Currently, web services are growing in an explosive speed. Many services that can perform same or similar functions arise. So, users should make choice among those services that can fulfill the given task. As more services can be used, users do not only care about the functional requirements but also the non-funtional requirements, such as service price, execution time, service availability. Therefore, how to select proper web services in composition as to satisfy the non-funtional restriction from users becomes a critical problem.To deal with this problem, firstly, we use QoS (Quality of Service) to describe the non-funtional properties for a single web service which forms an important basis in web serive selection. Then, we use the QoS model of compositive service to calculate the QoS for the entire process which can be used as the foundation of the fitness function in the algorithm.In this paper, we propose a service composition approach based on the genetic algorithm which treats the situation as a compostion optimization problem. Our approach adopts the evolution stabilization strategy to expand the standard gerenic algorithm, adds the mutation operator when coding the problem with chromosomes and uses the QoS as the basis to define the fitness function of chromosomes. Therefore, we can desribe the instance of QoS-aware service composition problem by using the chromosome and optimize the initial instances repeatly by using operations such as selection, crossing and variation. When the algorithm ends, we can get best solutions of the QoS aware service composition problem.In real-world applications, if users are not willing to spend a lot of time to get the best solution, they can change the ending condition in the algorithm to quickly get a satisfied one. The QoS aware service composition method based on generic algorithm do not need the objective functions and restricton condtions have linear features. Thus, our approach can deal with more kinds of restrictions in different forms and can be used in a much wider spectrum.