Dynamic configuration of optimal web services composition based on the quality

Nowadays, some special attributes of web services such as being independent from platform, reusability, having a loosely coupled architecture and the ability to compose together in contrast to other applications have made them as components enjoying the capability of supporting various changes. Therefore, most organizations tend to use these application components in order to provide their customers with better services. Furthermore, the use of a single service cannot meet the needs of most customers. Consequently, the issue of composing these services aimed at increasing their efficiency is one of the important issues. As the number of these services is increasing day to day, the process of selecting and composition appropriate web services in terms of their quality and users' need from among numerous single services having same functionality but different quality attributes such as cost, response time, reliability and availability has become one of the main challenges associated with composing these web services. To solve this problem, much works have been done so far in various ways. In recent years, most of these studies have used a variety of heuristic algorithms or a combination of them. The reason for this is the acceptable running time of these algorithms to find a solution close to the optimal solution. Accordingly, the proposed methodology for this research is based on particle swarm heuristic algorithm inspired by patterns of birds. The main reasons to choose this algorithm include its flexibility, less parameters, easy to implement and low cost. In contrast, the disadvantage of this algorithm is premature convergence that has been tried to be solved by adding two functions of Inertia Coefficient Adjustment and Particle Modification to main algorithm. Following the N iterations of the algorithm, the inertia coefficient adjustment function is called. Modifying the amount of inertia coefficient, the function will be able to control how the search is performed and to decrease the running time. As the running process continues, the second function used to modify the particles aimed at improving them is called following the M iterations of the algorithm. It tries to prevent algorithm being trapped in a local optimum. Applying this function, some web services on top particles which do not satisfy the desired quality features of users will be replaced by a number of services on the dataset. This is done on condition that the services on the dataset are better than services on the particle. If the algorithm gets trapped in a local optimum, this function makes the algorithm to perform the search in a new sample space. The efficiency of this method is evaluated in a simulated environment and is compared with both conventional birds and genetic algorithms. The results indicate that the proposed method is more effective than conventional algorithms in terms of running time and success rate as well as addressing the problem of being trapped in a local optimum.