Hybrid Replication Schemes of Processes in Energy-Efficient Server Clusters

In order to fault-tolerantly perform an application process in a server cluster, the process is replicated on multiple servers. Even if some replicas of the process is faulty, e.g. due to stop-fault of the host server, the process can be successfully performed as long as some number of replicas are operational. The more number of replicas of a process are performed, the more reliably and availably the process can be performed but the more electric energy is consumed. In this paper, we make clear what software components each replica of a process includes. We newly propose new replication schemes which are composed of a primary replica and more types of secondary replicas so that the total electric energy consumption can be reduced while the reliability and availability of each process can be increased. In the evaluation, we show the electric energy consumption of servers can be reduced in the replication schemes compared with the traditional active and passive replication schemes.