An intelligent agent for adaptive processor allocation in parallel databases

The use of software agents has become an increasingly valuable approach in dealing with information filtering and resource allocation. Parallel query execution has gained currency in high performance databases in recent years, and existing works have focused on either developing parallel algorithms for individual query operations or parallelising query execution plans, but few of them have studied processor allocation for parallel processing. This paper presents an intelligent agent system that exploits both intra-operation and inter-operation parallelism for multiple query execution. It monitors data loading with learning capabilities and processes queries adaptively based on phase-based and non-phase-based processor allocation strategy. An experimental study is presented to assess the performance benefits of the proposed agent system.