Energy-efficient Query Processing And Optimization With QoS Constraints

With the growing demands for data management, it is a great challenge to reduce and throttle the energy consumption in data centers. DBMS is a vital software in data center to deal with volumes of data. Thus, energy efficient query processing and optimization is one of the critical issues to be solved. Traditional DBMS only considers performance-oriented query optimizationt, while this thesis will take energy cost into account. The contributions are as follows:1. By evaluating the energy cost of each operator, we propose a novel energy cost model to estimate energy consumption of the whole query plan in different storage levels (Hard Disk, SSD, Memory Class). A dynamic programming method is used to search the optimal query plan for different optimization objectives.2. In light of a possible future storage memory, we propose an energy efficiency hybrid hash join algorithm under the QoS (Quality of Service) constraints. By adjusting the number of memory-resident partitions, a trade-off scheme between the time and energy consumption of the query plan is proposed.The experiments show that the performance-oriented optimization are consistent with the energy-effieient one under the traditional hardware architecture, but it’s different in the new one. The new energy cost model can lead to new query execution plans. By regulating the number of partitions located in the memory during the hybrid hash join execution, we can trade off between performance and energy-conservation under the QoS constraint.