Research On Key Problems In QoS-Based Web Service Composition

As a distributed application mode, the technology of services computing has obtained widespread application attention and system actualization in recent years. In the mode of services computing, the various heterogeneous resources (such as software, hardware, data and business) distributed on the Internet could be encapsulated into the form of Web services, and published through the unified interface and specification, which enables the resource providers to quickly deploy the outer business applications, according to users’personalized requirements. However, a single Web service is often limited in its functionality, and cannot satisfy the business requirements that are more and more complicated. In this situation, the Web service composition technology arises correspondingly. Through service composition, different Web services could be composed into a more powerful value-added service, so as to satisfy the users’complicated business requirements. Under this application background, the quality of service, i.e., QoS becomes an important basis for evaluating the application effect of service composition. QoS-based Web service composition has become a hot research topic in the domain of services computing, and attracts much attention from the academic and industrial areas.Currently, several problems and challenges are present in QoS-based Web service composition, which are mainly specified as follows:(1) for the large-scale Web service composition problem (i.e., the solution space of Web service composition problem is large), the present Web service composition methods are inefficient, in achieving a QoS optimal Web service composite solution; (2) the service composition environment is usually dynamic and unstable, therefore, a Web service may fail during its execution; in this situation, how to deal with this kind of service composition exception caused by Web service failure, so as to enhance the robustness of Web service composition, is a technique problem that needs urgent resolution; (3) when performing QoS evaluation, the traditional service selection methods challenge by some uncertain factors, due to the dynamic services computing characteristics, such as flexible QoS definition. In view of these problems, the main contribution of this paper is listed as follows:1. For the large-scale Web service composition problem, the low efficiency of present methods in deriving a QoS optimal Web service composite solution is analyzed. And a Web service composition method named LOEM (Local Optimization and Enumeration Method, LOEM) is brought forth, which is based on local QoS optimization. More specifically, in LOEM, on one hand, the candidate space of Web service composition is reduced by local QoS optimization, so that the efficiency of Web service composition is improved; on the other hand, for the obtained Web services after local QoS optimization, the IP (Integer Programming, IP) technique is employed to enumerate all the composite solutions, in order to seek a QoS near-to-optimal composite solution. Then through LOEM, a QoS near-to-optimal Web service composite solution could be found by service composition engine, within a shorter time period; and a better tradeoff is achieved, between the efficiency of Web service composition and the QoS optimality of the derived Web service composite solution.2. For the application challenge of service composition execution exception caused by Web service failure, the context factors of exception occurrence are analyzed; after which, an idea of exception handling is put forward by relaxing QoS constraints, as well as two principles for relaxing QoS constraints. Based on the two principles for relaxing QoS constraints, different QoS constraints are assigned with different step sizes and different order for relaxation. On this basis, an exception handling method named RelaxingMe (QoS constraints relaxing method, RelaxingMe) is put forward, which is based on the idea of relaxing QoS constraints. More specifically, in the proposed RelaxingMe method, by relaxing the overconstrained QoS constraints, an appropriate Web service is selected to replace the failed Web service, so as to resume the execution of Web service composition process, and enhance the robustness of Web service composition.3. The present Web service selection method is improved, so as to support the Web service selection with flexible QoS definition. More specifically, a QoS advertisement manner described by a value domain is brought forth, which supports service providers to realize the flexible QoS advertisements, so as to attract various users with different QoS preference. On this basis, a concept of dependency function is put forward, to depict the possible dependency relationship among various QoS values of an identical Web service, i.e., specify the flexible QoS levels of a Web service, from a more abstract point of view. Besides, during the QoS evaluation, a concept of satisfaction degree function is brought forth, to describe the variation trend of a user’s satisfaction with the variation of a service’s QoS value. Thus, during Web service selection, a user could specify his/her preferable and personalized QoS evaluation manner. With the dependency function and satisfaction degree function, a Web service selection method named FAEM (Flexible service quality Advertisement and Evaluation Method, FAEM) is proposed, which supports flexible QoS definition. FAEM not only faciliates the service providers for flexible QoS advertisements, but also helps the users to enact personalized QoS evaluation.4. Based on the Web service dataset QWS Dataset version 2.0, a set of simulated experiments are designed and developped. In the experiments, the Web service composition method LOEM and the exception handling method RelaxingMe are tested and verified, to evaluate their feasibility and effectiveness in dealing with their respective service composition problems. Besides, forand RelaxingMe, comparison verification and analysis are also made with present methods.