Cross Service Providers Workload Balancing for Data Centers in Deregulated Electricity Markets

Abstract

The emerging Internet of things and 5G applications boost a continuously increasing demand for data processing, which results in an enormous energy consumption of data centers (DCs). Considering that existing distributed geographical load balancing is approaching the limit in reducing the energy cost of DCs, cloud service providers (SPs) are motivated to pursue a higher level of cooperation. In this context, cross-SP workload balancing among the DCs operated by different SPs represents a future trend of the DC industry. This article investigates the optimal cross-SP workload balancing when it couples with the electricity markets. First, we assume that there is a central operator (CO) coordinating the DCs owned by various SPs. A noncooperative game is formulated to model the interaction between utilities and CO, which serves as a price maker. Under the centralized coordination of CO, an optimal solution is obtained with an iterative algorithm. Taking into account the computation and privacy issues, a decentralized algorithm is then proposed by utilizing techniques in a state-based potential game. Numerical results corroborate the effectiveness of the proposed algorithm. Simulations using Google workload trace show that the workload balancing among cross-SP DCs results in a lower DC operation cost than the existing price taker approach.