Capacity planning for data center to support green computing

Abstract

We propose a data center resource management framework to support green computing. This framework is composed of the power and workload management, and capacity planning schemes. While an action of power and workload management is performed in a short-term basis (e.g., fraction of minute), a decision of capacity planning is made in a long-term basis (e.g., few months). This paper mainly addresses a capacity planning problem. With the power and workload management, we formulate a capacity planning optimization as a stochastic programming model. The solution is the number of servers to be installed/deployed in a data center over multiple periods. The objective of this optimization model is to minimize the long-term cost under workload demand uncertainty. From the performance evaluation, with the proposed optimization model for the capacity planning scheme, the total cost to operate the data center in the long-term basis can be minimized while the job waiting time and job blocking probability are maintained below the target thresholds.

Sequential Learning-based IaaS Composition

“…Note, determining the optimal semantic levels, i.e., value ranges are out of the focus of this paper. We have added a new experiment to depict the effect of different semantic levels (𝑆𝑙𝑣 = [2,3,5,10,20]) on the accuracy of the proposed approach. The value ranges are uniformly distributed in the experiment.…”

“…The market-driven cloud service provisioning is a topical research area [60]. There exist several key service provisioning models in the cloud market such as on-demand, reservation [10,75], and economic models [47,55,58]. We propose an alternative strategy to the on-demand model which would be used in conjunction.…”

Capacity Planning for Green Data Center Sustainability

“…• Power and energy consumption through energy footprint from implemented virtualization that enables flexibility for capacity management in a data center [15] • To improve data center performance and availability from the maximum utilization of energy storage resources via an investigation into battery-aging affects data center sustainability. Green data center power costs and energy have been proven to reduce, and supply/demand power mismatch eliminated, with the use of energy storage devices [14] • Power and workload management for minimizing long-term cost due to uncertain workload demand is a capacity planning problem that is also being addressed [21] • Energy source selection (Green and Brown energy) for constructing a Green data center to meet cost, emission, and service availability requirement [13] Fig. 1 Standard Components for Data Center Capacity Planning [22] Research in capacity planning for data centers has been mainly focusing on the energy and cooling management perspective for data center optimization, as presented in Fig.…”