Multiprocessor Scheduling with Support by Genetic Algorithms-Based Learning Classifier System

The paper proposes using genetic algorithms - based learning classifier system (CS) to solve multiprocessor scheduling problem. After initial mapping tasks of a parallel program into processors of a parallel system, the agents associated with tasks perform migration to find an allocation providing the minimal execution time of the program. Decisions concerning agents' actions are produced by the CS, upon a presentation by an agent information about its current situation. Results of experimental study of the scheduler are presented.

[1]  Arjan J. C. van Gemund,et al.  LLB: A fast and effective scheduling algorithm for distributed-memory systems , 1999, Proceedings 13th International Parallel Processing Symposium and 10th Symposium on Parallel and Distributed Processing. IPPS/SPDP 1999.

[2]  Mohan Kumar,et al.  A flexible clustering and scheduling scheme for efficient parallel computation , 1999, Proceedings 13th International Parallel Processing Symposium and 10th Symposium on Parallel and Distributed Processing. IPPS/SPDP 1999.

[3]  Albert Y. Zomaya,et al.  Multiprocessor scheduling using mean-field annealing , 1998, Future Gener. Comput. Syst..

[4]  David E. Goldberg,et al.  Genetic Algorithms in Search Optimization and Machine Learning , 1988 .

[5]  Franciszek Seredynski,et al.  Scheduling tasks of a parallel program in two-processor systems with use of cellular automata , 1998, Future Gener. Comput. Syst..

[6]  Ian Glendinning,et al.  Parallel and Distributed Processing , 2001, Digital Image Analysis.

[7]  Tarek A. El-Ghazawi,et al.  A Parallel Genetic Algorithm for Task Mapping on Parallel Machines , 1999, IPPS/SPDP Workshops.

[8]  Ishfaq Ahmad,et al.  Dynamic Critical-Path Scheduling: An Effective Technique for Allocating Task Graphs to Multiprocessors , 1996, IEEE Trans. Parallel Distributed Syst..