Network-Aware Server Consolidation for Wireless Data Centers

Sayed Ashraf Mamun, Amlan Ganguly, Minseok Kwon, Andres Kwasinski, Panos P. Markopoulos · 2019

Servers in a data center are underutilized due to overprovisioning which contributes heavily toward the high-power consumption of the data centers. Server consolidation can drastically reduce the power consumption of data centers by engaging fewer servers and keeping the unused servers in deep sleep. However, server consolidation alone can adversely affect network performance. Recently, wireless data centers using high-speed millimeter-wave (mmWave) communications have been proposed and shown to drastically reduce the power consumption of the networking equipment. Therefore, server-consolidation strategies should leverage the novel data center network architectures to minimize power consumption with minimal impact on performance. Neither novel server consolidation algorithms for wireless data centers nor the impact of server consolidation on wireless data centers have been explored so far. In this paper, we propose a novel network-aware server consolidation algorithm called Bandwidth-Constrained Consolidation (BCC), for wireless data centers. Our case studies indicate that up to 37% power reduction can be achieved with the proposed consolidation algorithm in a wireless data center. A similar trend in power reduction is observed even with traditional wired data centers. However, the improvement in power consumption is higher in a wireless data center. Moreover, leveraging the network characteristics in the consolidation algorithm even improves the network performance for both wireless and wired data centers, preventing any adverse effect of consolidation on performance.

Read the paper · More papers on PaperTik