Energy aware database management

Abstract

Data center and cloud providers are responsible for providing services such as storage or retrieval for large amounts of (customer owned) data by using databsae management systems (DBMS). Service provision implies a specific quality of service regarding performance or security. Another factor of increasing importance is energy consumption. Although not a top priority for most customers, the cost of energy and thus (indirectly) the cost of service provision is key for both, customer and provider. Typically, energy consumption is viewed as a hardware related issue. Only recently, research has proved that software has a significant impact onto the energy consumption of a system too. Database management systems comprise various algorithms for efficiently retrieving and managing data. Typically, algorithm efficiency or performance is correlated with execution speed. This paper reports our results concerning the energy consumption of different implementations of sorting and join algorithms. We demonstrate that high performance algorithms often require more energy than slower ones. Furthermore, we show that dynamically exchanging algorithms at runtime results in a better throughput. © 2014 Springer-Verlag.