Adaptive load-balancing strategies for distributed systems
- 2 January 2003
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
Abstract
Describes SMALL, a system for learning load-balancing strategies in distributed computer systems. The load balancing problem is an ill-posed optimization problem because its objective function is ill-defined. Realistic state-space representations of this problem do not satisfy the Markov property. Experimentally feasible learning environments for load balancing exhibit delayed, evaluative feedback. Such aspects complicate the learning of strategies for load balancing. SMALL uses comparator neural networks for learning to compare objective-function values of states resulting from a set of alternative moves. The problem of learning from delayed evaluative feedback, also called the credit-assignment problem of reinforcement learning, is solved only for Markovian problems. The paper presents a novel credit-assignment procedure suitable for load balancing and other non-Markovian learning tasks.Keywords
This publication has 13 references indexed in Scilit:
- Imbedding gradient estimators in load balancing algorithmsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2003
- Physical-level synthetic workload generation for load-balancing experimentsPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2003
- Integrated Architectures for Learning, Planning, and Reacting Based on Approximating Dynamic ProgrammingPublished by Elsevier ,1990
- GAMMON: a load balancing strategy for local computer systems with multiaccess networksIEEE Transactions on Computers, 1989
- Learning to Predict by the Methods of Temporal DifferencesMachine Learning, 1988
- Strategy Learning with Multilayer Connectionist RepresentationsPublished by Elsevier ,1987
- Using stochastic learning automata for job scheduling in distributed processing systemsJournal of Parallel and Distributed Computing, 1986
- Pattern-recognizing stochastic learning automataIEEE Transactions on Systems, Man, and Cybernetics, 1985
- An Application of Bayesian Decision Theory to Decentralized Control of Job SchedulingIEEE Transactions on Computers, 1985
- A Unix-Based Local Computer Network with Load BalancingComputer, 1982