Research On Key Techniques Of Workflow For Grid Computing Environments

Grid computing is theGrid is regarded as the third generation Internet after traditional Internet and Web portals, and grid workflow is one of the most important research areas in grid computing. More and more powerful and coolaborative grid applications require tremendous resources, and they are more and more complicated which are constrained by temporal and resource relationship as the grid researches and grid infrastructure greatly advance. Grid workflow can conveniently construct, execute, manage and monitor grid application, and automate grid applications with great efficiency. Due to the dynamism, distribution, heterogeneity and autonomy of grid applications of grid, traditional workflow technology can't effectively solve the relative problems of grid environment. Grid workflow technologies should be reconsidered for grid and grid applications.Workflow model is the abstraction of workflow process, which is the foundation of workflow management. Layered modeling method has run through the grid workflow process and resource models to adopt the dynamical grid environment and grid application requirements. Grid workflow process model models a grid workflow application as a extended directed acyclic graph, which is more current and intelligible. The formalization of the model is defined to remedy the insufficiency of no formal description and accurate definition in the directed acyclic graph. In service-oriented grid workflow resource model, layered resources and resource-access mechanism aparts grid resources from grid workflow when defining the process model, which will be mapped dynamicly when executing, to enhance the flexibility of grid workflow.Role-based grid workflow adaptive scheduling model analyses related grid workflow tasks with directed acylic graph, configures task priorities according to different roles, and does the scheduling operations of grid workflow applications combined with the dynamic features of grid resources. The preferential scheduling of ramification can lead to intercurrent executing of more tasks, which not only make the most of grid resources, but also enhance the efficiencis of grid workflow applications. Another method to improve the performance of workflow scheduling bestly is minimizing the executing time of those tasks who locate on the way with the maximal executing time. The urgent tasks are are top-priority, and after a long waiting period normal tasks can upgrade to avoid starving to death. The integration of best-resource and load-banlance schedulings can satisfy the requirements of urgent tasks and give attention to the balanced and high utilization of grid resources.New requirements and challenges of workflow fault tolerance are brought forward in grid computing environment. The distributed heartbeat detecting mechanism can inspect large numbers of heterogeneous resources, and the layered fault detecing model can adjust and extend with dynamic resources. The strategy-based fault tolerance mechanism differentiates the executing performances and applying scopes of fault processing strategies, which can search out satisfied processing method with application requirement and resource status. All of these shield fault tolerance from grid workflow modeling, and are convenient for the efficient management and dynamic extension of the most strategies.Combining above research content, a grid workflow management prototype has been developed in ChinaGrid Image Processing Grid Platform, which support remote-sensing image processing applications, resource sharing and cooperation. It provides a flexible and dynamic image processing problem-solving environment. Grid workflow technology can facilitate agile requirements of various applications, general service broker can invoke heterogeneous grid services simply, and workflow models are defined by the more operable extended directed acyclic graph language. Remotely sensing applications have been applied on the platform successfully.