Research On Job Scheduling, Execution And Fault Tolerance On Optical Grid

As the ever increase of data communication and ever complicated jobs to execute, the implement of grid application becomes more and more challenging. In this case, advanced transferring network is indispensable, so raises the concept of Optical Grid, which interconnects grid resources with high speed optical network to accelerate the grid data exchange. In optical grid, specified scheduling is also important, while the complexly integrated optical grid environment requires high rank of fault tolerance.In this paper, we research the job scheduling, execution and fault tolerance mechanism, and propose a comprehensive infrastructure of optical grid scheduler to provide solution to the said problems. It employs job scheduling adapted by the scarcity of optical network resources to optimize the job execution and data transfer on optical grid. This scheduling schedules scarce optical network resources and grid resources jointly. At the acquirement of scheduling result, the scheduler proceeds to execute the jobs and create light-paths for data transfer between the jobs according to job execution process. What's more, to enable the accomplishment of jobs in case of failures of grid resources and optical network resources, the scheduler has robust and efficient fault tolerance mechanism. And we have also implemented this scheduler to support an aircraft design application, and conduct experimental evaluation of the whole system on optical grid test-bed. The application's practical running demonstrates its feasibility and high efficiency. Then we analyze the practical running scenarios of Optical Grid, and modify scheduling models to improve the conformity between scheduling and practical execution. Our work contributes to further research.