Fulfilling Task Dependence Gaps for Workflow Scheduling on Grids

The workflow programming paradigm has had a representative growth in the last years. This model is useful to represent flows of control and facilitate the complexity management of processes that have multiple dependent tasks. With the emergence of e-Science, workflow is becoming a standard for management of scientific processes with massive data sets. Within the workflow execution, scheduling of tasks is primordial to provide efficiency and to speed up the process results arrival. In this paper we consider the execution environment as being a computational grid, which is dynamic, non-dedicated, and has heterogeneous resources. We present a strategy for scheduling dependent task processes, dealing with scheduling and execution of more than one process at the same time potentially using resources in common. The algorithm is dynamic and adaptive, rescheduling tasks that are on the queue of resources not presenting good performance. Simulations show that the proposed strategy can give better schedules by enhancing the resources usage.