Towards Virtualization Deployment And Energy Efficient Database Cluster Design

With the explosive spread of cloud computing in 21 century, datacenter has been booming in all over the world. Never the less, the huge energy consumption and carbon dioxide emission caused by the development of the large scale datacenter, is becoming one of the biggest limit for datacenter. It’s more and more significant to reduce energy consumption of datacenter. We have no time to delay to researching on energy efficient and low emission datacenter. In this paper, we deploy our database cluster using virtual machines, thus designing a kind of heterogeneous cluster with virtual nodes. Based on this cluster, we design new data partitioning strategy, query execution strategy and virtual node deployment strategy.First of all, we definitely defined what the energy efficiency is in database cluster. Then, we analyze the energy consumption each part of the hardware contribute to, and propose a reasonable power model to manage energy consumption. After that, we design and implement a middleware for the virtualized database cluster, in order to carry out deeper research on energy saving of this cluster. Then, we propose 2 kind of data partitioning strategy, 5 kind of query execution strategy and 2 kind of virtual node deployment strategy to improve energy efficiency of the heterogeneous cluster.At last, we carry out a series of experiments to prove our power model, our strategy and our heterogeneous cluster is feasibility in this paper. We analyze the energy feature and the bottleneck of the cluster, introduce the direction of follow-up research.