Avaliação do impacto de estratégias de economia de energia em grades computacionais entre-pares.

Abstract

Peer-to-peer opportunistic grids are distributed computing infrastructures that harvest the idle computing cycles of computing resources geographically distributed. In these grids, the demand for resources is typically bursty. During bursts of resource demand, many grid resources are required, but on other times they remain idle for long periods. If the resources are kept powered on even when they are neither processing their owners workload nor grid jobs, their exploitation is not efficient in terms of energy consumption. One way to reduce the energy consumed in these idleness periods is to place the computers that comprise the grid in a “sleeping” mode which consumes less energy. In this work, we evaluated two sleeping strategies, namely: standby and hibernate. In grid computing, these strategies show a tradeoff between the benefit of energy saving and the associated costs in terms of increasing the job makespan and the effects in hard disks’ lifetime. The overhead in the makespan due to the time taken to wake up the resource when a new task arrives. The effect in hard disks’ lifetime is due to starting and stopping of the hard drives revolutions when sleeping strategies are used. In this work, we use a simulated model in order to evaluate this tradeoff. We also evaluated the minimum amount of machine idle time after which a sleeping strategy should be applied and how to choose which machine should be woken up, if several options are available. Our results show that both Standby and Hibernate strategies can present great energy savings. However, Standby presents lower impact on the job’s makespan. Furthermore, we have identified that sleeping strategies can be used as soon as the machine becomes idle, i.e., it is not necessary to wait any time in idle state. Regarding the strategies to chose machines, the ones that consider the machine energy efficiency increase the energy saving. Finally, we found that the use of sleeping strategies by peer-to-peer grids result in fewer hard disk transitions than would occur if the local user, instead of donate your machine to the grid, to adopt a strategy to place it in a sleeping mode, when it was idle.