Grid Workflow Key Technologies

More and more powerful computing and collaborative Grid applications that require tremendous resources are constructed as the Grid researches and Grid infrastructure greatly advance. Many applications are more and more complicated which are constrained by temporal and resource relationship. Grid workflow can conveniently construct, execute, manage and monitor grid applications, and automate grid applications with great efficiency. Due to the dynamism, distribution, heterogeneity and autonomy of grid applications, conventional workflow technology can't effectively solve the relative problems of grid environment. According to characteristics of grid resources and grid applications, the thesis studies grid workflow process model, organization and resource model, dynamic scheduling algorithm. The detail is as following:Grid workflow architecture. The architecture is composed of grid workflow user port, grid workflow service management, grid workflow organization and resource management, grid workflow performance monitoring and evaluation and grid workflow engine. The architecture complying with the standard of WFMC and GCC is open and extensible.Grid workflow process model. The dynamic workflow net named D-Petri Net is presented based on the WF-Net, which introduces the idea of dynamism and high level Petri. It can solve the problem of expansion, dynamism and multi-selection of grid process model. The grid process model constructed by D-Petri net can be analyzed and simulated by the existed petri net technology and tools according to the analysis and verification of relationship between D-Petri Net and Petri net.Grid workflow organization and resource model. The virtual organization in the grid environment is analyzed. The relationship of members in the virtual organization is defined. Based on the grid resource management model and virtual organization model. The integrated organization and resourced model is presented, and the algorithms of querying, adding, exiting about grid organization andresource are put forward. The integrated model can facilitate the dynamicalscheduling for workflow engine.Grid workflow dynamic scheduling algorithm. The grid tasks are classified. The grid tasks scheduling phases and policy are analyzed. The grid workflow dynamic scheduling algorithm with multi-policy is presented, which take consideration of the characteristics of grid workflow and grid resources. The scheduling of grid workflow tasks can be effectively implemented by the Key Factor, Dynamic Factor and Prior Factor.Grid workflow prototype and experiments. The prototype developed with the architecture and technology is introduced, which is constructed based on Globus Toolset. The experiment of genome sequencing based on shotgun algorithm is analyzed and the result is promising.