Game-Based Negotiation between Power Demand and Supply in Green Datacenters

Abstract

The power consumption of datacenters is growing rapidly and becoming a major concern. For reducing carbon footprint and increasing energy efficiency, a promising solution is to locally supply datacenters with renewable energies. However, a challenging problem in building such green datacenter is the coordinating between the power demand and the intermittent power supply. To address this problem, we propose to model the green datacenter as two subsystems, namely, Information Technology (IT) subsystem which consumes energy, and electrical subsystem which supplies energy. Then we aim to find an efficient compromise between the power supply and power demand, taking into account the constraints of both subsystems. Based on buyer-supplier game, we introduce a negotiation approach, in which the two subsystems are modeled as the energy buyer and energy supplier. A negotiation algorithm is proposed, allowing the two subsystems to negotiate and reach an efficient trade-off, while respecting their own utility/monetary gain. The algorithm is evaluated in our middleware of renewable energies-powered datacenter. The experimental results show that the proposed algorithm allows the negotiation process to reach stable points. This algorithm also obtains significant improvement in the datacenter's utility and quality of service (QoS), compared to the algorithms in which joint IT-energy management is not considered.