ABSTRACT
In performance engineering, metrics are often used to track the progress over time. Concerning the potential bias of using a single metric, performance engineers tend to use multiple metrics for reasoning. However, this approach has its own challenges. In this work we study one of the challenges in the context of analyzing trends in server energy proportionality. We examine a wide range of metrics for measuring energy proportionality, trying to determine which metrics are essential and which are redundant. We do this by comparing the trend curves of the metrics for the published results of the SPECpower_ssj2008 benchmark. While the context is specific, the proposed analysis method is quite general. We hope that this method would help us do performance engineering more effectively.
- L. Barroso and U. Hölzle. The case for energy-proportional computing. IEEE Computer, 40(12):33--37, Dec. 2007. Google ScholarDigital Library
- C. Belady. In the data center, power and cooling costs more than the IT equipment it supports. Electronics Cooling Magazine, Feb. 2007.Google Scholar
- L. Gray, A. Kumar, and H. Li. Workload characterization of the SPECpower ssj2008 benchmark. In SPEC International Performance Evaluation Workshop, June 2008. Google ScholarDigital Library
- The Green500 list: Environmentally responsible supercomputing. http://www.green500.org.Google Scholar
- C. Hsu, J. Kuehn, and S. Poole. Towards efficient supercomputing: Searching for the right efficiency metric. In International Conference on Performance Engineering, Apr. 2012. Google ScholarDigital Library
- C. Hsu and S. Poole. Revisiting server energy proportionality. In International Workshop on Power-Aware Systems and Architectures, Oct. 2013. Google ScholarDigital Library
- F. Ryckbosch, S. Polfliet, and L. Eeckhout. Trends in server energy proportionality. IEEE Computer, 44(9):69--72, Sept. 2011. Google ScholarDigital Library
- T. Scogland, C. Stefen, T. Wilde, F. Parent, S. Coghlan, N. Bates, W. Feng, and E. Strohmaier. A power-measurement methodology for large-scale, high-performance computing. In International Conference on Performance Engineering, Mar. 2014. Google ScholarDigital Library
- M. Stojmenović, A. Nayak, and J. Zunic. Measuring linearity of planar point sets. Pattern Recognition, 41(8):2503--2511, Aug. 2008. Google ScholarDigital Library
- B. Subramaniam and W. Feng. GBench: Benchmarking methodology for evaluating the energy efficiency of supercomputers. In International Supercomputing Conference, June 2012.Google Scholar
- B. Subramaniam and W. Feng. Towards energy-proportional computing for enterprise-class server workloads. In International Conference on Performance Engineering, Apr. 2013. Google ScholarDigital Library
- B. Subramaniam and W. Feng. Enabling efficient power provisioning for enterprise applications. In International Symposium on Cluster, Cloud and Grid Computing, May 2014.Google ScholarDigital Library
- G. Varsamopoulos, Z. Abbasi, and S. Gupta. Trends and effects of energy proportionality on server provisioning in data centers. In International Conference on High Performance Computing, Dec. 2010.Google ScholarCross Ref
- D. Wong and M. Annavaram. Scaling the energy proportionality wall with KnightShift. IEEE Micro, 33(3):28--37, May/June 2013. Google ScholarDigital Library
- D. Wong and M. Annavaram. Implications of high energy proportional servers on cluster-wide energy proportionality. In International Symposium on High-Performance Computer Architecture, Feb. 2014.Google ScholarCross Ref
Index Terms
- Measuring Server Energy Proportionality
Recommendations
Revisiting Server Energy Proportionality
ICPP '13: Proceedings of the 2013 42nd International Conference on Parallel ProcessingServer energy proportionality refers to a proposed ideal that a server consumes energy proportional to its utilization level. Since its introduction in 2007, the ideal has been adopted by the server industry as a design goal to further optimize the ...
Scaling the Energy Proportionality Wall with KnightShift
Measuring energy proportionality accurately and understanding the reasons for disproportionality are critical first steps in designing future energy-efficient servers. This article presents two metrics—linear deviation and proportionality gap—that let ...
Trends in Server Energy Proportionality
Server energy proportionality, as quantified by the proposed EP metric, has improved significantly, from 30-40 percent in 2007 to 50-80 percent today, but much more can be done to move systems closer to ideal.
Comments