Energy Proportional Servers: Where Are We in 2016?
作者： Congfeng Jiang , Yumei Wang , Dongyang Ou , Bing Luo , Weisong Shi
关键词: Efficient energy use 、 Real-time computing 、 Server 、 Energy (signal processing) 、 The Internet 、 Data center 、 Energy accounting 、 Computer science 、 Energy proportional computing 、 Energy consumption 、 Distributed computing 、 Workload
摘要: The huge energy consumption in data centers produces not only high electricity bill but also tremendous carbon footprints. Although today's servers and of leading internet companies are more efficient than ever before, the fluctuations external workload internal resource utilization calls for proportional computing. Insight into server proportionality can help improve placement while reducing consumption. In this paper, we investigate all 477 valid published results SPECpower_ssj benchmark from 2007 to 2016Q3 reorganize them by hardware availability year accurate analysis on production servers. Through comprehensive find that: (1) specious stagnation recent years is mainly caused adoption processors specific microarchitecture indicative trend improvement. (2) Microarchitecture evolution has influence efficiency improvement proportionality. (3) Today's servers' peak efficiencies shifting 100% 80% or 70% improves with such shifting. We then conduct extensive experiments 4 rack variations under different configurations, including memory per core installation processor frequency scaling. Our show that configuration significant impact server's efficiency. findings presented paper provide useful insights guidance system designers, as well center operators aware savings.
sci-hub.se 参考文章(47)
Cullen Bash, Manish Marwah, Niraj Tolia, Zhikui Wang, Xiaoyun Zhu, Parthasarathy Ranganathan, Delivering energy proportionality with non energy-proportional systems: optimizing the ensemble international conference on cluster computing. pp. 2- 2 ,(2008)
Larry D Gray, Anil Kumar, Harry H Li, None, Workload Characterization of the SPECpower_ssj2008 Benchmark spec international performance evaluation workshop. pp. 262- 282 ,(2008) , 10.1007/978-3-540-69814-2_17
Daniel Wong, Julia Chen, Murali Annavaram, A Retrospective Look Back on the Road Towards Energy Proportionality ieee international symposium on workload characterization. pp. 110- 111 ,(2015) , 10.1109/IISWC.2015.18
Klaus-Dieter Lange, Identifying Shades of Green: The SPECpower Benchmarks IEEE Computer. ,vol. 42, pp. 95- 97 ,(2009) , 10.1109/MC.2009.84
Daniel Wong, Murali Annavaram, Scaling the Energy Proportionality Wall with KnightShift IEEE Micro. ,vol. 33, pp. 28- 37 ,(2013) , 10.1109/MM.2013.31
Balaji Subramaniam, Wu-chun Feng, Towards energy-proportional computing for enterprise-class server workloads international conference on performance engineering. pp. 15- 26 ,(2013) , 10.1145/2479871.2479878
Qingyuan Deng, David Meisner, Abhishek Bhattacharjee, Thomas F. Wenisch, Ricardo Bianchini, CoScale: Coordinating CPU and Memory System DVFS in Server Systems international symposium on microarchitecture. pp. 143- 154 ,(2012) , 10.1109/MICRO.2012.22
Wei Deng, Fangming Liu, Hai Jin, Bo Li, Dan Li, Harnessing renewable energy in cloud datacenters: opportunities and challenges IEEE Network. ,vol. 28, pp. 48- 55 ,(2014) , 10.1109/MNET.2014.6724106
Daniel Wong, Murali Annavaram, KnightShift: Scaling the Energy Proportionality Wall through Server-Level Heterogeneity international symposium on microarchitecture. pp. 119- 130 ,(2012) , 10.1109/MICRO.2012.20
Rahul Singh, Prateek Sharma, David Irwin, Prashant Shenoy, K.K. Ramakrishnan, Here Today, Gone Tomorrow: Exploiting Transient Servers in Datacenters IEEE Internet Computing. ,vol. 18, pp. 22- 29 ,(2014) , 10.1109/MIC.2014.40