Genetic Algorithm On Web Services Selection Supporting Global QoS Constraints

How to create robust service compositions becomes the next step work in web services and has attracted a lot of researches. At the same time, web services selection supporting global QoS constraints plays an important role in web services composition. Since web services with the same functions and different QoS are increasing with the proliferation of web services, and web services requesters always express their functional requirements as well as their QoS constraints set, it is needed to select the best composite plan from numerous plans in order to maximize user satisfaction and satisfy the consumers' global QoS constraints. The researches about practicability, effectivity, stability and adaptive capability of selection mechanism have gained considerable momentums.In order to resolve services selection with the global QoS constraints, a novel genetic algorithm characterized by fine stability, quick convergence rate and practicability is presented. It is the genetic algorithm with population diversity handling mechanism. The following are the main points:1. In the literatures, the presented genetic algorithms always adopted one dimension coding scheme that can not represent effectively composite service. In order to resolve this problem, a special relation matrix coding scheme of chromosomes is presented. It can express simultaneously all of composite paths, which can not be expressed simultaneously by the one dimension coding scheme. This matrix can also represent effectively the composite service re-planning and cyclic paths with the help of a simple method. Many composition scenarios can also be showed by the matrix but not by the one dimension. By running only once, the genetic algorithm with the proposed coding scheme can construct the composite service plan according with the QoS requirements from a great deal of services compositions with different QoSes.2. According to the evolution theory based on natural selection of Darwin, the species evolve in the form of the whole group but not individuals. A population diversity handling mechanism is presented to control the population evolution. It enables the population to evolve on the basis of the whole population evolution principle of the biologic genetic theory. Prematurity is overcome effectively through the conservation of the historical optimal population and the competition between the historical optimal population and the current population.3. The characteristic of randomicity ensures that genetic algorithm has the ability to search all of paths of web services composition, but it also introduces slow convergence, great differences among results after running many times, a soaring overhead along with increasing composite size. Aiming at solving these problems, some policies are proposed in order to direct the evolution of genetic algorithm. They are an enhanced initial population policy and an evolution policy. They ground GA in the practicability on web services selection with global QoS constraints.4. In order to resolve services composition with the global QoS constraints, a framework of QoS-aware services composition is presented. The core of the framework is a novel middleware (GAMi) for QoS-aware web services selection. According to the global QoS constraints of services composition proposed by users, the middleware can get the optimal composition from many composition candidates that accord with the functional requirements. The GAMi also can categorize the users and do different actions on the basis of the classes of the users. Meanwhile, some flow charts about the services composition and the services selection are presented.