Research On Service Scheduling Mechanism In Grid

Based on Service Oriented Architecture, the grid, which characterized by its capacity for coordinating resource sharing across multiple organizations and providing problem solving environment for virtual organization, has been evolving with enterprise computing in the field of goals, ideas, implementations etc to the new grid environment—Service Grid. The fundamentality of service grid is facilitating service sharing and collaborating to afford on demand service. Therefore, it is necessary to have a scheduling mechanism to solve the problem of automatic service composition, dynamic service discovery and matching, runtime service binding etc.The characteristics of grid such as dynamic, heterogeneous and autonomous make it more difficulty to implement grid scheduler than local resource scheduler. For example, grid scheduler needs to deal with resource sharing across domain and distributed security etc. Moreover, also there are differences between implement mechanism of scheduling in service grid and computing grid. For in service grid, the resources are not only naked hardware resources which mainly used for batch scientific computing tasks in computing grid, but also including varied resources encapsulated by grid services that need frequently interactive access. Further, the evaluating parameters for service scheduling are extended to multiple constraints such as Service Level Agreement (SLA), Quality of Service (QoS), Quality of Experience (QoE) etc.At present, the researches on grid scheduling are mainly fall into three aspects: the organized model of grid resource, the model of grid application and grid scheduling algorithm. The paper emphasizes the technologies for resource organizing model, scheduling model and implement mechanism of scheduling in service grid and distinguishes the key point for realizing service scheduling. By combining Petri Net and workflow technologies, a formal service scheduling process model is presented. Further, two multiple QoS constraints service scheduling algorithms are presented and a scheduling system—G-Sched was implemented. The main contributions of this paper as follows:(1) Semantic service based description and organizing model are presented for supporting dynamic service scheduling.By employing ontology service, Open grid service architecture is extended to support semantic service. At the same time, the core specification of OGSA-WSRF(Web Service Resource Framework) are detailed about how to model service resource. The description and organization of WS-Resource is enhanced with semantic by extends Web Service Description Language metadata model.(2) A dynamic service scheduling model is discussedThe key technologies of grid service scheduling are analyzed and Color-Petri Net based grid service scheduling model are presented with which combining semantic description model of WS-Resource, the technologies of workflow and dynamic web service composition. The proposed model characterized by integrating exception handle, data transfer and QoS constraint that has been analyzed and verified.(3) Two service scheduling algorithm under multiple QoS constraints are proposedBased on the investigation of the existing grid application model and scheduling problem from the view of dynamic service composition, state-chart based grid service composition model is constructed, which transform the problem of grid service scheduling to the problem of service search and matching by coordinating flow-based web service composition with dynamic QoS constraints. Two heuristic algorithms for solving the problem of service scheduling under multiple QoS constraint are proposed and verified efficiency through experiment and simulation.(4) The grid service scheduling component is implementedThe dynamic grid service scheduling process and involved key technologies are analyzed and the...