Data center power reduction by heuristic variation-aware server placement and chassis consolidation

The growth in number of data centers and its power consumption costs in recent years, along with ever increasing process variation in nanometer technologies emphasizes the need to incorporate variation-aware power reduction strategies in early design stages. Moreover, since the power characteristics of identically manufactured servers vary in the presence of process variation, their position in the data center should be optimally determined. In this paper, we introduce two heuristic variation-aware server placement algorithm based on power characteristic of servers and heat recirculation model of data center. In the next step, we utilize an Integer Linear Programming (ILP) based variation-aware chassis consolidation technique to assign tasks to servers such that total power consumption of data center is minimized under obtained server placement. Evaluating the proposed server placement heuristics reveals that up to 14.85% power savings can be achieved under different utilization rates compared to conventional random server placement method.