An energy-efficient power management for heterogeneous servers in data centers

Abstract

Power management for heterogeneous servers has been playing a key role in improving energy efficiency in data centers. Running latency-critical web services on such scenario is still challenging due to the overheads of task transition between such servers. In this paper, we present a runtime power management system, Montgolfier, which is built on a latency-aware feedback control mechanism. It consolidates wimpy and brawny servers into composite nodes performing latency-critical applications to improve overall energy efficiency while ensuring QoS. The key idea behind Montgolfier is to mitigate the negative effect of server switches by dynamic load prediction and to determine thin-provisioned configurations in fine-grain manner within servers for daily fluctuating loads. Our evaluation results show that Montgolfier reduces energy consumption by up to 34.9% without violating any QoS constraints.