An Online Algorithm for Power-proportional Data Centers with Switching Cost
share
Recent works show that power-proportional data centers can save energy cost by dynamically adjusting active servers based on real time workload. The data center activates servers when the workload increases and transfers servers to sleep mode during periods of low load. In this paper, we investigate the right-sizing problem with heterogeneous data centers including various operational cost and switching cost to find the optimal number of active servers in a online setting. We propose an online regularization algorithm which always achieves a better competitive ratio compared to the greedy algorithm in lin2012online. We further extend the model by introducing the switching cost offset and propose another online regularization algorithm with performance guarantee. Simulations based on real world traces show that our algorithms both outperform the greedy algorithm.
READ FULL TEXT