Improving energy efficiency in colocation data centers for demand response

摘要

Colocation data centers (colocations, for short) are developing rapidly and have become ideal participants for emergency demand response (EDR) programs. However, even colocation operators wish to save energy, they cannot achieve energy reduction without coordination from the tenants, because the servers in the colocations are owned and operated by the tenants. To solve the ?uncoordinated relationship? issue between operators and tenants, some works have been done by way of incentivizing the tenants to reduce the energy consumption of their servers. However, apart from servers, the power consumption of the cooling system also accounts for a large portion in a colocation, which should be optimized as well. Unfortunately, the servers and the cooling system are controlled by tenants and operators separately in colocations, and thus ?uncoordinated relationship? issue also exists between IT and cooling systems. In this paper, coarse-grained and fine-grained incentive mechanisms are proposed to solve these problems. Approximation algorithms are developed to optimize the energy-saving problems. Furthermore, Vickrey?Clarke?Groves (VCG) theory is introduced into our incentive mechanism design to guarantee the feasibility and truthfulness of the two mechanisms. Trace-driven simulations are performed to validate the effectiveness of the two incentive mechanisms. The results show that compared with the existing incentive mechanisms, up to 20.50% of the energy-saving cost can be reduced in the coarse-grained mechanism and 28.34% of the cost reduction can be achieved in the fine-grained mechanism.