Dynamic Right-Sizing for Power-Proportional Data Centers

TLDR

Power consumption imposes a significant cost for data centers implementing cloud services, yet much of that power is used to maintain excess service capacity during periods of low load. The study examines how much power can be saved by dynamically right‑sizing data centers through server shutdowns during low‑load periods and how to realize this via an online algorithm. We propose a general model, prove that the optimal offline algorithm has a simple reverse‑time structure, and use this to design a new lazy online algorithm that is 3‑competitive, contrasting it with receding horizon control. Validation on two real data‑center workload traces demonstrates that the lazy online algorithm can achieve significant cost savings.

Abstract

Power consumption imposes a significant cost for data centers implementing cloud services, yet much of that power is used to maintain excess service capacity during periods of low load. This paper investigates how much can be saved by dynamically “right-sizing” the data center by turning off servers during such periods and how to achieve that saving via an online algorithm. We propose a very general model and prove that the optimal offline algorithm for dynamic right-sizing has a simple structure when viewed in reverse time, and this structure is exploited to develop a new “lazy” online algorithm, which is proven to be 3-competitive. We validate the algorithm using traces from two real data-center workloads and show that significant cost savings are possible. Additionally, we contrast this new algorithm with the more traditional approach of receding horizon control.