A green scheduler for cloud data centers using renewable energy

In recent years, a large number of cloud data centers have been built around the world to provide all kinds of service. The high power consumption of the servers in those data centers deteriorates the already serious global warming. Many Internet operators have devoted much effort in power-aware scheduling using renewable resources. However, the intermittent nature of renewable resources proposes new challenges for us: how to distribute the incoming requests to the data centers that are powered by renewable energy, while minimizing the carbon emissions under a certain budget. In this paper, we model the problem as a constraint optimization problem. The goal is to minimize total carbon emissions of those geographically distributed data centers, while satisfying the constraints: (1) The electricity budget in each time slot; (2) The intermittent supply of renewable energy; (3) The number of each type of heterogeneous servers in each data center; (4) QoS constraint in request processing time. We ingeniously transformed this problem into a mixed integer linear programming problem, and solved it using Cplex. Our simulation is based on traces from real world, and we can get optimum solution when problem scale like Wiki workload is not so large. Experiments show that our scheduler can minimize carbon emissions while satisfying the above mentioned constraints.