Skip to Main Content
In this paper, we analyze the performance, utilization, and power estimation of server systems by both adopting the batch service and adjusting the batch size. In addition to reducing system utilization and power consumption the batch service can also maintain an adjustable system performance. When few events are accumulated in the queue of the batch service, a longer waiting time is needed to reach the threshold value for starting the batch service. Therefore we add a time counter to use in conjunction with the batch service in addition to an event counter; this time counter effectively reduces the mean system response time. But now the clock rate of developed server systems is becoming faster. Adopting the batch service is regarded as the system mechanism for processing events, but it may affect a system's performance and power consumption. We use the queueing model method to analyze the system state and to learn the variations among power consumption, system utilization and mean response time by adjusting the size of the batch service. Furthermore we can evaluate how to adjust a batch size via the derived equation in order to obtain a suitable balance among performance, utilization and power estimation for a server system.
Date of Conference: 19-21 Nov. 2007