Global cost diversity aware scheduling algorithm for heterogeneous data centers

Internet based companies service user requests from multiple data centers located around the globe. These data centers often house heterogeneous computing infrastructures and draw electricity from the local electricity market. Electricity is also consumed for cooling the data centers, and the total costs often run into millions of dollars. Reducing operating cost, therefore, is an important and challenging problem. In this work, we propose a novel solution which schedules user requests to geographically diverse data centers and exploits the server heterogeneity, data center power usage efficiency, and global electricity market diversity to reduce operating costs. We evaluate our solution using a real-world workload, electricity prices, and power efficiency values. We show that our scheduling algorithm achieves a cost savings up to 14% over a load balancing scheme that distributes requests evenly across data centers, and outperform existing solutions which do not exploit either electricity market diversity or data center hardware diversity.