Dynamic Clusters of Servers for Reducing Electric Power in P2P Overlay Networks

Abstract

Electric power consumed by servers has to be reduced in order to realize green societies. For each service request from a client, a server is selected in a cluster of servers. Here, the total electric power consumption of the servers has to be reduced since servers mainly consume electronic power. We consider computation (CP) and storage (ST) types of application processes performed on servers in this paper, where CPU and storage drives are mainly used, respectively. In our previous studies, the energy-aware (EA) algorithm is discussed to select a server in a static cluster of servers for performing each request process so that the total power consumption of the servers can be reduced. However, an idle server consumes electric power even if no process is performed and an idle server consumes electric power at more than half of the maximum power consumption rate. It is critical to discuss how to reduce electric power consumed by idle servers in addition to energy-efficiently performing processes on servers. In this paper, a cluster is composed of servers which support some service like computation and storage service. Each cluster takes servers in a pool of idle servers distributed on networks so that only servers required to perform request processes but no idle server is included in each cluster. The membership of a cluster is dynamically changed depending on the traffic. Servers are selected in a server pool and join a cluster if the traffic increases in the cluster. On the other hand, idle servers leave a cluster for the server pool if the traffic decreases. We evaluate dynamic clusters in terms of the total power consumption and average execution time.