Differentiating the Starting Time of Processes in Energy-Efficient Redundant Execution Algorithm

Application processes have to be not only efficiently and reliably but also energy-efficiently performed in presence of server faults in distributed applications. In the improved redundant power consumption laxity-based (IRPCLB) algorithm proposed in our previous studies, a request of a process is sent to multiple servers. Then, replicas of a process are redundantly performed on more than one server. Here, the reliability and availability of the cluster can increase. However, the larger amount of electric power is consumed than non-redundant execution. Once a replica of a process terminates on a server, meaningless replicas on the other servers are forced to terminate in order to reduce the total power consumption. In this paper, we newly discuss an extended IRPCLB (EIRPCLB) algorithm to furthermore reduce the total power consumption of a server cluster by differentiating the starting time of each replica. Here, a process request is not broadcast and is rather serially sent to each server every delta time units. We evaluate the EIRPCLB algorithm in terms of total power consumption of a cluster, the average response time and number of servers to perform each process, and the operating ratio of a cluster compared with the IRPCLB algorithm. We show how to decide on the inter-request time delta for each request given the requirements on maximum response time and maximum number of active servers.