MTS: Multiresolution Thread Selection for Parallel Workload Distribution

Authors:
Chonglei Mei;Hai Jiang;Jeff Jenness
Affiliations:
Department of Computer Science, Arkansas State University, Jonesboro, Arkansas, USA 72467;Department of Computer Science, Arkansas State University, Jonesboro, Arkansas, USA 72467;Department of Computer Science, Arkansas State University, Jonesboro, Arkansas, USA 72467
Venue:
GPC '09 Proceedings of the 4th International Conference on Advances in Grid and Pervasive Computing
Year:
2009

Citing 5
Cited 0

Grid Services for Distributed System Integration

Computer
Arachne: A Portable Threads System Supporting Migrant Threads on Heterogeneous Network Farms

IEEE Transactions on Parallel and Distributed Systems
Heterogeneous Process Migration: The Tui System

Heterogeneous Process Migration: The Tui System
Process/Thread Migration and Checkpointing in Heterogeneous Distributed Systems

HICSS '04 Proceedings of the Proceedings of the 37th Annual Hawaii International Conference on System Sciences (HICSS'04) - Track 9 - Volume 9
Dynamic Scheduling with Process Migration

CCGRID '07 Proceedings of the Seventh IEEE International Symposium on Cluster Computing and the Grid

Quantified Score

Hi-index	0.00

Visualization

Abstract

Computing workload distribution is indispensable for resource sharing, cycle stealing and other modes of interaction in distributed systems/Grids. Computations should be arranged to adapt the capacity variation of system resources. Although computation migration is the essential mechanism to move computing tasks around, the decision making of which task should be relocated is even more critical, especially when multithreaded parallel programs are involved. Multiple threads might be treated as partial workload and moved together. Based on thread similarity, this paper proposes a novel Multiresolution Thread Grouping algorithm (MTG) to classify threads into hierarchical Thread Bundles (TB) some of which can be picked by Multiresolution Thread Selection scheme (MTS) for load distribution. During the process of MTG, global variables are reorded so that one-time migration cost and post-migration communication volume and frequency can be reduced. Experimental results demonstrate the effectiveness of MTS for parallel workload distribution.