An energy-efficient process migration approach to reducing electric energy consumption in a cluster of servers

In order to realise eco society, we have to reduce electric energy consumed by servers in a cluster. Clients issue requests of application processes to servers in a cluster. Here, application processes have to be efficiently performed on servers with respect to not only performance but also energy consumption. In this paper, we newly propose an energy-efficient process migration (EMG) approach to energy-efficiently performing application processes on servers. Here, processes on servers migrate to other servers so that the processes are more efficiently performed. We first discuss how to estimate the termination time of each current process on a server and the total electric energy consumption of a server to perform every current process. A process performed on a current server migrates to another server if the server is expected to consume smaller electric energy to perform the process than the current server and the deadline constraint on the process is satisfied on the server by using the estimation models. In the evaluation, we show the total energy consumption of servers is smaller and the average execution time of each process is shorter in the EMG algorithm than non-migration algorithms like the round robin and random algorithms.