Skip to Main Content
Most of the previous research on data centers power consumption has focused on understanding how to minimize the power consumption inside the data center. It is, however, also important to investigate the power consumption associated with transporting data between data centers and end users. In this paper, we consider three problems. First, through linear programming (LP) models and through simulations we determine the optimal location of a data center or multiple data centers in an IP over wavelength-division-multiplexing network so as to minimize the network's power consumption. Here, we consider the impact of network topology, traffic profile, upload and download rates, number of data centers and the impact of power minimization on delay. Second, we study how to replicate content that has different popularity to minimize power consumption through the use of an LP model. Here, we consider five classes (but the models are general) of content that have different levels of popularity and consider multiple data centers. The optimization attempts to identify where to store a data object that has a given popularity such that the network's power consumption is minimized. We have also developed a novel routing algorithm, energy-delay optimal routing, to minimize the power consumption of the network under replication while maintaining QoS. Third, we investigate through LP the problem of whether to locate data centers next to renewable energy or to transmit renewable energy to data centers in a given network topology under different traffic conditions and taking into account the network components' power consumption. Given a number of wind Farms whose locations are known together with the electrical power transmission losses, we identify the optimal location of data centers such that the network's power consumption is minimized and consider a network where the nodes that are not connected to wind farms have access to solar power. The results show that by identifying the- - optimum data center locations, combining the multi-hop bypass heuristic with renewable energy and the replication scheme, power consumption savings of up to 73% can be achieved.
Date of Publication: June15, 2011