Research On Distributed Reliable Service Composition Mechanism In Pervasive Environment

Pervasive computing, which develops from distributed computing and mobile computing, is a new generation of computing technology. Compared with traditional web service composition, since pervasive environment has attribute of mobility and heterogeneity, pervasive service composition is often more likely to fail. To deal with this relatively frail service environment, it is necessary to establish a reliable service composition mechanism in pervasive environment.To address the problems above, after doing some researches on attributes of pervasive computing and other graph-based service composition methods, this paper proposes an improved service composition method based on graph. In execution phase, this method allows to rollback to a previous service's output state when a failure occurs and start a compensatory composition plan. To deal with the possible failure of a service and strengthen fault-tolerance, this paper further proposes a "prevention" and "test - compensation" tactic, then gives three reliable algorithms using the improved service composition method based on graph in a simple "single-engine" system structure, and forms a single-engine based reliable service composition mechanism in pervasive computing. At last, this paper gives three corresponding reliable algorithms in "multi-engines" hierarchical structure, and proposes a distributed reliable service composition mechanism in pervasive computing.This paper brings the following innovations:(1)To deal with the service's failure-handling which is not supported by existing graph-based service composition, an improved service composition method based on graph is proposed. The method contains two stages to generate and execute the service composition plan, and introduces the support for rollback to the execution phase. If a failure occurs when calling a service, the method allows rollback to a previous service's output state, then starts a new service composition plan to replace the failure parts of the original composition plan.(2)For the service composition's generating and executing stages, a "prevention" and "test - compensation" tactic is proposed to improve service composition's reliability. Based on this tactic, three reliable algorithms are proposed in "single-engine" system structure. The first algorithm generates a basal service composition plan, and then analyzes each component service's reliability. The second algorithm generates a backup service composition plan for those component services with lower reliability. The third algorithm generates a compensation plan with the lowest cost for failed service in execution stage.(3)According to the difference of pervasive devices' duties and functions, this paper puts forward three device roles, including: service composition engine, service execution proxy and service provider. Based on these roles, this paper divides the pervasive devices into three levels: 0-2, and proposed a "Three-Level and Four-Layer" hierarchical structure with its structuring protocol.(4)Considering the structure characteristics of the pervasive environment, this paper extends the reliable algorithms in (2) to "Three-Level Four-Layer" hierarchical structure, and ultimately, proposes a distributed reliable service composition mechanism in pervasive computing.The experimental data shows that the above reliability algorithms greatly improve the reliability of service composition.