Applying Probabilistic Adaptation to Improve the Efficiency of Intra-Query Load Balancing

Abstract

In the context of adaptive query processing (AQP), several techniques have been proposed for dynamically adapting/redistributing processor load assignments throughout a computation to take account of varying resource capabilities. The effectiveness of these techniques depends heavily on when and to what they adapt processor load assignments, particularly in the presence of varying load imbalance. Most existing approaches to this problem use heuristics based only upon the current machine load levels. The authors provide an algorithm, prAdapt that probabilistically predicts the future load on processors, based upon the recent history. It uses this prediction to evaluate the expected performance of different alternative solutions, taking into account the cost of the adaptation itself. If it finds a better solution than the current load distribution policy, it adapts to that distribution. Using a simulation based evaluation; they compare prAdapt to other approaches for AQP reported in the literature. The authors’ simulation results indicate that prAdapt often outperforms these other approaches.