A General Distributed Scalable Peer to Peer Scheduler for Mixed Tasks in Grids

Abstract

We consider non-preemptively scheduling a bag of independent mixed tasks in computational grids. We construct a novel Generalized Distributed Scheduler (GDS) for tasks with different priorities and deadlines. Tasks are ranked based upon priority and deadline and scheduled. Tasks are shuffled to earlier points to pack the schedule and create fault tolerance. Dispatching is based upon task-resource matching and accounts for computation as well as communication capacities. Simulation results demonstrate that with respect to the number of high-priority tasks meeting deadlines, GDS outperforms prior approaches by over 40% without degrading schedulability of other tasks. Indeed, with respect to the total number of schedulable tasks meeting deadlines, GDS outperforms them by 4%. The complexity of GDS is O(n2m) where n is the number of tasks and m the number of machines. GDS successfully schedules tasks with hard deadlines in a mix of soft and firm tasks, without a knowledge of a complete state of the grid. This way it helps open the grid and makes it amenable for commercialization.