F. Dong and S.G. Akl (Canada)
Task Scheduling, Grid Computing, DAG
Resources in the computational Grid nowadays are usually owned by multiple administrative domains which are connected by WANs. Grid resources have their own local management policies and a Grid scheduler cannot control resources beyond the domain of itself. Scheduling algorithms designed for traditional systems do not consider these new characteristics in Grids. In this paper, a two-phase scheduling approach for DAG-based applications in the Grid is proposed. In the first phase, the task graph is partitioned according to the status of selected available resource clusters by an algorithm called GPRC. In the second phase, an algorithm called HEFT is used to map task nodes to processing nodes inside a resource cluster. GPRC does not require detailed status information or control privilege on every Grid resource so that the dependence on Grid information services is reduced and, at the same time, the higher priority of local resource management policies is respected. Analysis and experiment show the effectiveness and adaptation of this new approach in the Grid scenario.
Important Links:
Go Back
