Research On Service Composition Method Based On Physical Programming

Practical and applied systems are faced with requirement change during every stage such as design and operation. Although the problem can be solved by improving the existing system or replacing it with a new system, it tends to take the users great manpower and financial resources. In order to achieve good inheritance of existing systems and cut down users’expenditures, Service-Oriented Architecture (SOA) emerges as the times require. SOA encapsulates different applications into services, which can contact each other through well-defined interfaces and composite into new comprehensive services according to user requirements. In addition to the traditional large enterprise software, new applications such as Internet of Things, Ubiquitous Networks and Cloud computing also need the support of SOA. They have brought some new problems and technical challenges for SOA, including service discovery, service composition modeling, service selection optimization, etc.This dissertation focuses on the service composition problem. The user preferences are considered during the modeling of service composition, which is achieved by a modified physical programming method based on weight function. Following the characteristic analysis of mobile Ad hoc networks (MANET), a load balancing service discovery protocol is proposed to find candidate services for building composite service. Simultaneously, a modified heuristic cross entropy algorithm is proposed to optimize the service selection problem. Furthermore, through the development of heterogeneous terminals and gateway, a multi-terminal collaboration experiment platform for service composition is constructed to verify the feasibility and effectiveness of the proposed service discovery protocol, service composition model and service selection algorithm. The concrete contents of this dissertation are given in details as follows:1. To make physical programming more satisfied with the OVO rule, a weight function-based physical programming (WFPP) is proposed. Firstly, all the design metrics to be optimized are normalized to the same range. Then a weight function for each design metric is defined, which endows the design metric with a verified weight. Finally, the aggregate objective function is formulated by the weighted sum of all design metrics. WFPP is applied in both MOO problems and generating Pareto solutions. Simulation results demonstrate that WFPP is effective in these applications while satisfying the user preference.2. Since MANET is dynamic, resource-constrained and usually performs as a temporary group, a service discovery protocol based on local maximal independent subset cache is proposed to find multiple candidate services for service composition. The protocol follows a reactive cashing policy, that is, the nodes cache the service information only when a service request or response is received, In order to balancing the network load, the service cache nodes are selected by local maximal independent subset method. The simulation results demonstrate that the proposed protocol outperforms traditional protocols in terms of response delay, message load and cache load.3. WFPP is adopted to model the service composition problem. Two kinds of user preferences are integrated into the modeling process:the preference ranges that express user’s satisfaction to each objective, and the fuzzy importance that presents the relative importance between all the objectives. Simulation results indicate that the optimal solution is well balanced between various objects, and the objects with higher importance have priority to be optimized. In addition, since the service selection is a NP-hard problem, a modified heuristic cross entropy algorithm is proposed to optimize it. The algorithm reduces the computational complexity of service selection by deleting and sorting the candidate services according to the local optimal evaluation index. The effectiveness of the proposed algorithm is verified in some experiments.4. According to theoretical research, a multi-terminal collaboration experiment platform is proposed to implement service composition. The platform demonstrates the communication of heterogenous networks, service discovery of heterogenous terminals, and service composition based on user preferences, providing a verification of the proposed key technologies.