Optimizing query execution to improve the energy efficiency of database management systems

Increasing energy costs became one of the critical issues in database centers in the recent years. The consciousness to turn towards energy-preserving technologies have put concepts like power-awareness into the spotlight. But especially databases lack the capability of managing the energy consumption while operating and past research solely focussed on improving performance characteristics. This master’s thesis investigates potential modifications of the postgreSQL query optimizer and executor to improve the overall energy efficiency of the query processing engine. For this, this study introduces an energy cost model on the optimizer level and the use of dynamic voltage and frequency scaling on the executor level. Additional concepts like deadlines are implemented to exploit the full range of functions and they are subsequently combined and used to maximize the positive impact of the algorithms designed for energy efficiency on the energy consumption. The extended framework is capable of reducing active energy costs significantly which is proven by applying the TPC-H benchmark and a concluding discussion of additional extensions including query scheduling shows that even further energy savings can be achieved.