An optimization framework for data centers to minimize electric bill under day-ahead dynamic energy prices while providing regulation services

Abstract

Considering the growing number of Internet and cloud computing data centers in operation today and the high, yet flexible data center electric load, data centers can be good candidates to offer ancillary services and respond to regulation signals in a smart grid. This paper considers a problem whereby the smart grid employs both day-ahead dynamic energy prices and regulation signals to incentivize (cloud) data centers to simultaneously reduce their energy consumptions and participate in an ancillary service market. A data center controller schedules task dispatch and performs resource allocation in order to minimize the overall cost, which is the total electricity cost based on time-of-use energy prices minus any monetary compensations that data center may receive due to offering ancillary services. Moreover, the data center must satisfy average latency requirements in processing requests as specified in service-level agreements with clients. A two-tier hierarchical solution is presented for the data center controller, which achieves optimality in minimizing the overall cost with polynomial time complexity. Experimental results on Google trace demonstrate the effectiveness of the proposed solution in minimizing the overall cost in the data center.

Evaluation process of demand response compensation models for data centers

“…Benefits for the DC range from the increase of grid reliability through outage avoidance to various options for the reduction of electricity cost. In terms of minimizing electricity cost in DCs this is elaborately discussed in [6] [11]. For price-based DR, this is achieved by considering the current electricity price and adjustment of the load accordingly [15].…”

Distributed control system for demand response by servers

“…Others use collocated batteries as additional storage devices in a data center [13,14]. Still others use day-ahead pricing signals or other stochastic devices for predicting server workload or load imbalance on the grid [15,16].…”

Fast demand response with datacenter loads

“…This suggests the possibility that the "spare capacity" of computer servers can be used to provide load-shaping services to the electric grid and indeed this idea has been proposed [5], and is being actively studied by researchers [4,[6][7][8][9][10][11][12][13].…”

“…There has been significant recent interest in the idea of datacenters as a demand response resources [4,[6][7][8][9][10][11][12][13]].…”

“…For example, the authors of [7] presented an economically-driven algorithm for distributed workload migration in a demand response context. The proposal [9] made an economic case for for datacenter demand response, providing a mathematical framework for optimizing datacenter workload against day-ahead dynamic pricing. At faster timescales, [11] presents simulation results and a control strategy for the joint optimization of datacenter and plug-in electric vehicles for frequency regulation.…”