Scheduling Divisible Loads in Parallel and Distributed Systems
Scheduling Divisible Loads in Parallel and Distributed Systems
Bandwidth-Centric Allocation of Independent Tasks on Heterogeneous Platforms
IPDPS '02 Proceedings of the 16th International Parallel and Distributed Processing Symposium
Autonomous Protocols for Bandwidth-Centric Scheduling of Independent-Task Applications
IPDPS '03 Proceedings of the 17th International Symposium on Parallel and Distributed Processing
Stork: Making Data Placement a First Class Citizen in the Grid
ICDCS '04 Proceedings of the 24th International Conference on Distributed Computing Systems (ICDCS'04)
A-FAST: autonomous flow approach to scheduling tasks
HiPC'04 Proceedings of the 11th international conference on High Performance Computing
Hi-index | 0.00 |
In this paper, we address the problem of building and maintaining dynamic overlay networks on top of physical networks for the autonomous scheduling of divisible load Grid applications. While autonomous scheduling protocols exist to maximize steady-state throughputs for given overlay networks, not much work has been done on building the most efficient overlay. In our work, nodes use the bandwidth-centric principle to select other nodes with which overlay edges must be formed dynamically. The node which has the entire dataset initially (the scheduler) starts forming the overlay and the nodes which receive tasks at rates greater than their task execution rates further expand it. We use simulation studies to illustrate the functioning of our overlay forming mechanism, and its robustness to changes in the characteristics of the system resources.