An Adaptive Hash Join Algorithm using Mobile Agents
Jean-Paul Arcangeli, Abdelkader Hameurlain, Frédéric Migeon, Franck Morvan · 2002
In this paper, we propose a decentralized model for dynamic optimization of hash join operator well adapted to wide-area distributed databases, and its implementation using mobile autonomous agents. Agents' autonomy allows the execution plan to react dynamically to estimation errors without being penalized by time overheads that usually results from the use of a centralized monitor. Performance evaluation shows the benefits of this approach for standard hash join algorithms. Response time can significantly be reduced : from 3% to 60% when the error on selectivity factor exceeds 40%, and from 11% to 400% when estimation error on the size of a temporary relation is greater than 30%. Performance results serves as a basis for the definition of a strategy for agent mobility.