Minimizing Environmental footprints of data centers under budget and service requirement constraints

The energy consumption of data centers (DCs) has been increasing, which will continue due to the increase of Internet traffic and stringent service level agreements (SLAs). Analogously, the protection of global and local environments has also driven the regulation authorities to encourage energy consumers, especially corporate entities, for the usage of green energy sources. However, the green energy is usually more expensive (up to four to five times for some cases) than the traditional energy generated from coal and petroleum. One essential problem for managing DCs, according to the greenness tendency, is to minimize the environmental penalty (or equivalently to maximize the greenness) by dispatching the requests to proper DCs under the SLA and budget constraints. This paper presents optimization techniques for dynamic workload balancing for cloud-scale data center (DC) management. We present a model for commonly found electricity tariffs for green energy and provide an efficient heuristic algorithm to maximize its usage while incorporating its intermittent availability. We evaluate the presented solution with real-life traces of electricity prices and DC workloads. Extensive evaluations support our solution's potential to minimize the environmental penalty for Internet service providers under the budget while fulfilling their SLAs.