ABSTRACT
Dynamic voltage and frequency scaling (DVFS) is a mechanism adopted by major hardware vendors to reduce power demand during times of low processor utilization. However, reducing processor frequency to decrease power demand usually results in degraded services' performance leading to service level agreement violations. Governors, which are a piece of software at kernel level, are devised to exploit the flexibility provided by DVFS technologies of the hardware. Utilization-based governors change frequency and voltage at discrete time instances based on workload's utilization without taking into account performance constraints of services. In this paper, a model for the utilization-based Conservative governor is proposed. The model allows us to predict both service performance (mean response time) and processor power demand. An M/M/1 simulator is presented which is used to validate the accuracy of the proposed model. For model accuracy validation, a second methodology based on the frequency probabilities of the processor is proposed. Both approaches confirm the derived DTMC model. We also carry out a comparison between On-demand and Conservative governors and show that the latter performs better for Markovian workloads.
- 2007. Power provisioning for a warehouse-sized computer. In Proceedings of the 34th annual international symposium on Computer architecture (ISCA '07). ACM, New York, NY, USA, 13--23. Google ScholarDigital Library
- AMD. 2014. Cool'n'Quiet Technology @ONLINE. (Jan. 2014). http://www.amd.com/us/products/technologies/cool-n-quiet/Pages/cool-n-quiet.aspxGoogle Scholar
- Robert Basmadjian, Florian Niedermeier, and Hermann de Meer. 2016. Modelling Performance and Power Consumption of Utilisation-based DVFS Using M/M/1 Queues. In Proceedings of the Seventh International Conference on Future Energy Systems (e-Energy '16). ACM, New York, NY, USA, Article 14, 11 pages. Google ScholarDigital Library
- D. Bertsekas and R. Gallager. 1992. Data Networks. Prentice Hall. Google ScholarDigital Library
- Yuan Chen, Subu Iyer, Xue Liu, Dejan Milojicic, and Akhil Sahai. 2007. SLA decomposition: Translating service level objectives to system level thresholds. In Autonomic Computing, 2007. ICAC'07. Fourth International Conference on. IEEE, 3--3. Google ScholarDigital Library
- Joshua Dennis Booth, Jagadish Kotra, Hui Zhao, Mahmut T. Kandemir, and Padma Raghavan. 2015. Phase Detection with Hidden Markov Models for DVFS on Many-Core Processors. In 2015 IEEE 35th International Conference on Distributed Computing Systems. IEEE.Google Scholar
- Chen-Ying Hsieh, Jurn-Gyu Park, Nikil D. Dutt, and Sung-Soo Lim. 2015. Memory-aware cooperative CPU-GPU DVFS governor for mobile games.. In ESTImedia. IEEE, 1--8. http://dblp.uni-trier.de/db/conf/estimedia/estimedia2015.html#HsiehPDL15Google Scholar
- Enhanced Intel. 2004. SpeedStep® Technology for the Intel® Pentium® M Processor White Paper, March 2004. Technical Report. Recovered 30/1/2011 from World Wide Web: ftp://download. intel. com/design/network/papers/30117401. pdf.Google Scholar
- Shin-Gyu Kim, Hyeonsang Eom, Heon Y. Yeom, and Sang Lyul Min. 2014. Energy-centric DVFS Controlling Method for Multi-core Platforms. Computing 96, 12 (Dec. 2014), 1163--1177. Google ScholarDigital Library
- Etienne Le Sueur and Gernot Heiser. 2010. Dynamic Voltage and Frequency Scaling: The Laws of Diminishing Returns. In Proceedings of the 2010 International Conference on Power Aware Computing and Systems (HotPower'10). USENIX Association, Berkeley, CA, USA, 1--8. http://dl.acm.org/citation.cfm?id=1924920.1924921 Google ScholarDigital Library
- David Meisner, Brian T. Gold, and Thomas F. Wenisch. 2009. PowerNap: eliminating server idle power. In Proceedings of the 14th international conference on Architectural support for programming languages and operating systems (ASPLOS XIV). ACM, New York, NY, USA, 205--216. Google ScholarDigital Library
- Jean-Marc Pierson and Henri Casanova. 2011. On the Utility of DVFS for Power-aware Job Placement in Clusters. In Proceedings of the 17th International Conference on Parallel Processing - Volume Part I (Euro-Par'11). Springer-Verlag, Berlin, Heidelberg, 255--266. http://dl.acm.org/citation.cfm?id=2033345.2033372 Google ScholarDigital Library
- Akhil Sahai, Anna Durante, and Vijay Machiraju. 2002. Towards automated SLA management for web services. Hewlett-Packard Research Report HPL-2001-310 (R. 1) (2002).Google Scholar
- Kisho S. Trivedi and Robin Sahner. 2009. SHARPE at the Age of Twenty Two. SIGMETRICS Perform. Eval. Rev. 36, 4 (March 2009), 52--57. Google ScholarDigital Library
- Linlin Wu, Saurabh Kumar Garg, and Rajkumar Buyya. 2011. Sla-based resource allocation for software as a service provider (saas) in cloud computing environments. In Cluster, Cloud and Grid Computing (CCGrid), 2011 11th IEEE/ACM International Symposium on. IEEE, 195--204. Google ScholarDigital Library
Index Terms
- Modelling and Analysing Conservative Governor of DVFS-enabled Processors
Recommendations
Modelling performance and power consumption of utilisation-based DVFS using M/M/1 queues
e-Energy '16: Proceedings of the Seventh International Conference on Future Energy SystemsDynamic Voltage and Frequency Scaling (DVFS) has become a key technique to reduce power consumption at times of low processor (CPU) utilisation. Reducing CPU frequency usually results in degraded services' performance leading to Service Level Agreement (...
Workload-adaptive process tuning strategy for power-efficient multi-core processors
ISLPED '10: Proceedings of the 16th ACM/IEEE international symposium on Low power electronics and designAs more devices are integrated with technology scaling, reducing the power consumption of both high-performance and low-power processors has become the first-class design constraint. Reducing power consumption while satisfying required performance is ...
DVFS Governor for HPC: Higher, Faster, Greener
PDP '15: Proceedings of the 2015 23rd Euromicro International Conference on Parallel, Distributed, and Network-Based ProcessingIn High Performance Computing, being respectful of the environment is usually secondary compared to performance: The faster, the better. As Exascale computing is in the spotlight, electric power concerns arise as current exascale projects might need too ...
Comments