Research On Key Technology Of SOA Oriented Service Load Management

SOA(Service Oriented Architecture) is now the main software architecture of application system. With the continuously growth of functions and businesses, the loading pressure of services on SOA rise rapidly. To effectively cope with the huge service request pressure, it is widely used to deploy multiple concrete service instances to provide an abstract service. Heterogeneous of concrete services, different demand of business process for service quality and concrete services’distributed deployment result SOA oriented load management become more complex. Make effectively management of multiple concrete services has great significance to promote service performance and availability.To meet the demands of business process for service quality, make effectively manage of service resource, through in-depth analysis of the load management research status, this paper has made a research on key technology of SOA oriented framework of service load management, service instance selection and optimization method and quality of service assurance. Main content are as follow:1.Established a SOA oriented load management framework for periodic load distribution. Through analyze on the strengths and weakness of existing modes of front-end and back-end load management, by using system analysis and design, a kind of load management framework combine front and back-end control was put forward. The framework adopt a method of the front-end periodic assign service instance which is meeting the business process’demand of service quality, and the back-end eliminate the constrain violations of service instance by making load diffusion on neighbor instances. The method proposed effectively make the advantages of the combination of two modes, eliminate the single mode’s shortcomings to some degree, make the business processes’demand is met and promote the back-end service instances’performance and efficiency.2.Proposed a genetic algorithm based service instance selection and optimization method for back-end load migration. Directed at the request of business process for service quality, combine the performance indicator of back-end service instance, apply the Poisson combination and random distribute theorem, analyze and propose two optimization objectives, establish a multi-objective programming with nonlinear constrains. The genetic algorithm based on priority goal programming is used to solve the multi-objective optimization problem, achieving the business process to the service instance’initial deployment. Under the premise of meeting the business process for service quality constrains, full consideration is made on the back-end service instances of load and resources utilization, promote the total performance of service instances.3.Proposed a dynamic load balancing diffusion algorithm with the ability of neighbor loading awareness. Directed to the original diffusion algorithm has the problem of cannot aware of neighbor loading and load flow directivity is not strong, the first layer neighbor load information is used to improve the convergence rate of load balancing, this method is effectively to make up the shortcoming of the front-end distribution which has less control on runtime system.4.Designed and implemented a SimEvents based service load management system simulation framework.The design of key modules such as business process,service instance, service selection and optimization, service instances’ quality assurance are given. By the corresponding simulation experiment, it has proved that the framework can make effectively load management, satisfy the service quality request of business process and provide service quality assurance.