Study On The Resource Requirement Description Method And Services Selection Policy In Service Grid

With the rapid development of the Internet economy, grid services defined by the Open Grid Service Infrastructure (OGSI), are emerging as the extension of Web services that follows special conventions to allow the services and the resources it uses to be more easily shared by different constituencies. Grid services provide better interoperability and improve distributed system integration. In the business-oriented service grid system, different companies can connect their business applications with grid services, which help companies efficiently compositing services and reduce running expenses. With the rapid increase of grid services, to provide suitable grid services for companies on their requirements has become a very important problem in the service grid system.In this paper, we propose a service grid system, called Enterprise Flexible Composition Service (EFCS), for companies to dynamically compose their business grid services. In the system, a service selection module is designed to provide additional services including service filtering, service selection and dispatching, which ensure Grid services perform smoothly in a well-conducted and coordinated fashion.To provide the resource services that are more suitable for user's jobs, a resource service requirement description language is presented to help service consumers describe their requirement. In order to define the functional description, the description language combines the characteristics of Web Ontology language (OWL) and semantic constraint description. And with the Service Level Agreement (SLA) provided in the description language, selection of grid service is not only based on availability, but also based on Quality of Service (QoS) characteristics. The resource service requirement description language is integrated into the Grid Process Description Language (GPDL), with which user can describe their functionality requirement, semantic constraint, QoS requirement and business process of jobs.In this paper we propose a new model of service evaluation based on QoS to support the service selection policies, which is used in the service selection module. In order to sort candidate services of grid process by comparing their QoS characteristics, three service evaluation algorithms are designed in this model. We also introduce clustering analogy method into algorithms to analyze service candidates with QoS parameters. With the model, SLA and selection rules are set in the service selection policies to select a grid service that best fits user's requirement from sorted service candidates and guarantee QoS of service implement. We first evaluate the design performance of service selection module by comparing remote invoking times with different successful rate and different number of candidate service in module. The result shows it can improve the performance of grid system. Then we apply the evaluation algorithms to different grid services, and point out the...