Scheduling and Executing Heterogeneous Task Graph in Grid Computing Environment

摘要

In this paper, we propose a directed hypergraph with parameters to describe general, heterogeneous, distributed and dynamic grid task requests. With the design of GHDL (Grid Heterogeneous task graph Definition Language), grid task requests are uniformly encapsulated and recognized by heterogeneous computers. Furthermore, we provide an effective solution for scheduling parallel tasks with priority constraints and data dependency. With self-adaptive scheduling strategy library, different types of tasks can be executed and optimized. We deal with the heterogeneity of grid environment and tasks by matching client with hosting machine. The dynamic resource problem is solved by runtime monitoring and task migration. A simulation experiment of tree growth in grid environment shows very promising results.