Operating Cost Reduction for Distributed Internet Data Centers

摘要

In order to provide low latency, pay-as-you-go infrastructure, Cloud service provider has built a large number of geo-distributed data centers. Nowadays, energy cost has become increasingly important fraction of data center operating cost. Researchers' attentions are attracted to propose many methods to reduce the electricity cost for data centers. However, another significant operating cost, Internet bandwidth cost, has been neglected. Because both the electricity prices and bandwidth prices are varied with time and regions, it is not easy to minimize the operating cost while response to workload timely. As a solution, based on Lyapunov optimization framework, we propose an online control policy to achieve close-to-optimal performance with tradeoff between cost and delay. The advantage of our solution is that we do not need any future knowledge on the stochastic model (e.g., bandwidth prices and electricity prices). The results from our simulations on real workload trace and price data demonstrated that our solution is effective.