Cutting down the energy cost of geographically distributed cloud data centers

Abstract

The energy costs constitute a significant portion of the total cost of cloud providers. The major cloud data centers are often geographically distributed, and this brings an opportunity to minimize their energy cost. In this work, we model a geographically distributed data center network that is specialized to run batch jobs. Taking into account the spatio-temporal variation in the electricity prices and the outside weather temperature, we model the problem of minimizing the energy cost as a linear programming problem. We propose various heuristic solutions for the problem. Our simulations using real-life workload traces and electricity prices demonstrate that the proposed heuristics can considerably decrease the total energy cost of geographically distributed cloud data centers, compared to a baseline technique.