Title :
Reconfiguring multiprocessor systems while minimizing disturbance
Author :
Trotter, John ; Varvarigou, Theodora A.
Author_Institution :
AT&T Bell Labs., Murray Hill, NJ, USA
Abstract :
In this paper we consider the problem of reassigning tasks that were running on a failed processor of a multiprocessor system to the remaining fault-free processors. The new assignment of the tasks to the fault-free processors should 1) preserve the capacity constraints of the processors, 2) balance the load on the processors, 3) assign tasks to processors for which they have large affinity, and 4) minimize the disturbance caused by tasks that are moved from one processor to another during reassignment. It is easy to see that the problem above is NP-complete since it is a more general case of the bin-packing problem, for which many efficient heuristics and approximation algorithms have been proposed in the literature. In our work we propose new, efficient, heuristic algorithms for a problem more general than the bin-packing problem stated above, that takes all the above additional parameters into consideration
Keywords :
computational complexity; heuristic programming; multiprocessing systems; processor scheduling; reconfigurable architectures; NP-complete; approximation algorithms; bin-packing problem; capacity constraints; disturbance minimisation; heuristic algorithms; heuristics; multiprocessor systems reconfiguration; reassigning tasks; Application software; Approximation algorithms; Costs; Fault tolerant systems; Heuristic algorithms; Legged locomotion; Multiprocessing systems; NP-complete problem; Particle measurements; Polynomials;
Conference_Titel :
Fault-Tolerant Parallel and Distributed Systems, 1994., Proceedings of IEEE Workshop on
Conference_Location :
College Station, TX
Print_ISBN :
0-8186-6807-5
DOI :
10.1109/FTPDS.1994.494483