Project Details

WattDB is a clustered scale-out database system running on lightweight, amdahl-balanced nodes. In contrast to other distributed DBMSs, WattDB is not only capable of dynamically scaling out to additional nodes for increased performance, but also able to scale-in to reduce power consumption.

Our prototype is based on the engine we developped in the context of the SIGMOD 2010 Programming Contest whose goal has been set to come up with a distributed database engine. We have expanded its functionality towards adaptivity and energy efficiency. As every node is small and light-weight and, thus, consumes little energy, we have at least energy proportionality at the granularity of nodes. Depending on the load situation, nodes can be switched on and off, so this approach will approximate the ideal energy-proportional system. We believe that small-scale distributed systems are the key concept to achieve energy proportionality. By applying ad-hoc adaptivity mechanisms, energy consumption will scale with the given load.

In order to turn off nodes, their data has to migrate to other nodes first. Therefore, flexible partitioning schemes are needed to enable quick adapation to changing workloads. By distributing query operators among nodes, utilization can be balanced among all nodes in the cluster quickly.

In this project, we are exploring techniques to reduce energy consumption of query processing, data storage and adaptivity. Our vision is an energy proportional DBMS, that dynamically scales its number of nodes to satisfy the current workload. We have several publications on various aspects of the system, e.g. measuring power consumption, adapting the storage and processing layer as well as research papers focusing on the database system as a whole.