PREDICTION-BASED DYNAMIC LOAD-SHARING HEURISTICS

This paper presents dynamic load-sharing heuristics which are novel in that they use predicted resource requirements of processes to manage workload in a distributed system. A previously developed statistical pattern-recognition method is employed for resource prediction. While nonprediction based heuristics depend on rapidly changing system status (e.g., load levels), the new heuristics depend on slowly changing program resource usage patterns. Furthermore prediction-based heuristics can be more effective since they use ''future'' requirements rather than just current system state. Four prediction-based heuristics, two centralized and two distributed, are presented here. Using trace driven simulations, they are compared against random scheduling and two effective nonprediction based heuristics. Results show that the prediction-based, centralized heuristics achieve up to 30% better response time than the nonprediction, centralized heuristic, and that the prediction-based, distributed heuristics achieve even better (up to 50%) improvement relative to their nonprediction counterpart.