On-Line Task Migration in Hypercubes Through Double Disjoint Paths

Authors:
Hsing-Lung Chen;Nian-Feng Tzeng
Affiliations:
National Taiwan Institute of Technology, Taipei, Taiwan;Univ. of Southwestern Louisiana, Lafayette
Venue:
IEEE Transactions on Computers - Special issue on mobile computing
Year:
1997

Citing 5
Cited 1

Task migration in hypercube multiprocessors

ISCA '89 Proceedings of the 16th annual international symposium on Computer architecture
Constructing Parallel Paths Between Two Subcubes

IEEE Transactions on Computers
A Survey of Wormhole Routing Techniques in Direct Networks

Computer
Experimentation with Hypercube Database Engines

IEEE Micro
Job Scheduling is More Important than Processor Allocation for Hypercube Computers

IEEE Transactions on Parallel and Distributed Systems

Design and performance evaluation of combined first-fit task allocation and migration strategies in mesh multiprocessor systems

Parallel Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

Repeated subcube allocation and deallocation in hypercubes tend to cause fragmentation, which can be taken care of by task migration. Earlier task migration dealt with the establishment of a single path from each participating node for transmitting migrated information. The time required for migration with single paths is long, if a large amount of information is moved in hypercubes. This paper considers speedy task migration in that two disjoint paths are created between every pair of corresponding nodes for delivering migrated information simultaneously, reducing the size of data transmitted over a path. All migration paths selected are pairwise disjoint and contain no link of active subcubes, so that task migration can be performed quickly and on-line without interrupting the execution of other jobs. Our approach could lead to a considerable savings in the migration time for contemporary hypercube systems, where circuit switching or wormhole routing is implemented.