Optimal power allocation in server farms

ABSTRACT

Server farms today consume more than 1.5% of the total electricity in the U.S. at a cost of nearly $4.5 billion. Given the rising cost of energy, many industries are now seeking solutions for how to best make use of their available power. An important question which arises in this context is how to distribute available power among servers in a server farm so as to get maximum performance.

By giving more power to a server, one can get higher server frequency (speed). Hence it is commonly believed that, for a given power budget, performance can be maximized by operating servers at their highest power levels. However, it is also conceivable that one might prefer to run servers at their lowest power levels, which allows more servers to be turned on for a given power budget. To fully understand the effect of power allocation on performance in a server farm with a fixed power budget, we introduce a queueing theoretic model, which allows us to predict the optimal power allocation in a variety of scenarios. Results are verified via extensive experiments on an IBM BladeCenter.

We find that the optimal power allocation varies for different scenarios. In particular, it is not always optimal to run servers at their maximum power levels. There are scenarios where it might be optimal to run servers at their lowest power levels or at some intermediate power levels. Our analysis shows that the optimal power allocation is non-obvious and depends on many factors such as the power-to-frequency relationship in the processors, the arrival rate of jobs, the maximum server frequency, the lowest attainable server frequency and the server farm configuration. Furthermore, our theoretical model allows us to explore more general settings than we can implement, including arbitrarily large server farms and different power-to-frequency curves. Importantly, we show that the optimal power allocation can significantly improve server farm performance, by a factor of typically 1.4 and as much as a factor of 5 in some cases.

REFERENCES

• 1: Lesswatts.org: Race to idle. http://www.lesswatts.org/projects/applications-power-management/race-to-idle.php.
• 2: Intel: Nehalem. http://intel.wingateweb.com/US08/published/sessions/NGMS001/SF08_NGMS001_100t.pdf.
• 3: U.S. Environmental Protection Agency. Epa report on server and data center energy efficiency. 2007.
• 4: National Electrical Contractors Association. Data centers -- meeting today's demand. 2007.
• 5: Jeffrey S. Chase , Darrell C. Anderson , Prachi N. Thakar , Amin M. Vahdat , Ronald P. Doyle, Managing energy and server resources in hosting centers, Proceedings of the eighteenth ACM symposium on Operating systems principles, October 21-24, 2001, Banff, Alberta, Canada
• 6: Intel Corp. Intel Core2 Duo Mobile Processor Datasheet: Table 20. http://download.intel.com/design/mobile/datashts/32012001.pdf, 2008.
• 7: Mootaz Elnozahy , Michael Kistler , Ramakrishnan Rajamony, Energy conservation policies for web servers, Proceedings of the 4th conference on USENIX Symposium on Internet Technologies and Systems, p.8-8, March 26-28, 2003, Seattle, WA
• 8: Xiaobo Fan , Wolf-Dietrich Weber , Luiz Andre Barroso, Power provisioning for a warehouse-sized computer, ACM SIGARCH Computer Architecture News, v.35 n.2, May 2007
• 9: Wes Felter , Karthick Rajamani , Tom Keller , Cosmin Rusu, A performance-conserving approach for reducing peak power consumption in server systems, Proceedings of the 19th annual international conference on Supercomputing, June 20-22, 2005, Cambridge, Massachusetts [doi>10.1145/1088149.1088188]
• 10: Mark E. Femal , Vincent W. Freeh, Boosting Data Center Performance Through Non-Uniform Power Allocation, Proceedings of the Second International Conference on Automatic Computing, p.250-261, June 13-16, 2005 [doi>10.1109/ICAC.2005.17]
• 11: M. S. Floyd , S. Ghiasi , T. W. Keller , K. Rajamani , F. L. Rawson , J. C. Rubio , M. S. Ware, System power management support in the IBM POWER6 microprocessor, IBM Journal of Research and Development, v.51 n.6, p.733-746, November 2007
• 12: Anshul Gandhi, Mor Harcol-Balter, Rajarshi Das, and Charles Lefurgy. Optimal power allocation in server farms. Technical Report CMU-CS-09-113, 2009.
• 13: Intel Corp. Intel Math Kernel Library 10.0 -- LINPACK. http://www.intel.com/cd/software/products/asmo-na/eng/266857.htm.
• 14: Raj Jain. phThe Art of Computer Systems Performance Analysis: techniques for experimental design, measurement, simulation, and modeling. pages 563--567. Wiley, 1991.
• 15: Radim Kolar. Web bench. http://home.tiscali.cz:8080/cz210552/webbench.html.
• 16: Kleinrock L. Queueing Systems, Volume 2. Wiley-Interscience, New York, 1976.
• 17: Charles Lefurgy , Xiaorui Wang , Malcolm Ware, Power capping: a prelude to power shifting, Cluster Computing, v.11 n.2, p.183-195, June 2008 [doi>10.1007/s10586-007-0045-4]
• 18: J.D. McCalpin. Stream: Sustainable memory bandwidth in high performance computers. http://www.cs.virginia.edu/stream/.
• 19: David Mosberger , Tai Jin, httperf—a tool for measuring web server performance, ACM SIGMETRICS Performance Evaluation Review, v.26 n.3, p.31-37, Dec. 1998 [doi>10.1145/306225.306235]
• 20: Vivek Pandey, W. Jiang, Y. Zhou, and R. Bianchini. DMA-Aware Memory Energy Management. HPCA '06: The 12th International Symposium on High-Performance Computer Architecture, pages 133--144, 11-15 Feb. 2006.
• 21: Ramya Raghavendra , Parthasarathy Ranganathan , Vanish Talwar , Zhikui Wang , Xiaoyun Zhu, No "power" struggles: coordinated multi-level power management for the data center, Proceedings of the 13th international conference on Architectural support for programming languages and operating systems, March 01-05, 2008, Seattle, WA, USA
• 22: K. Rajamani, H. Hanson, J.C. Rubio, S. Ghiasi, and F.L. Rawson. Online power and performance estimation for dynamic power management. Research Report RC-24007, July 2006.
• 23: Salvatore Sanfilippo. WBox HTTP testing tool (Version 4). http://www.hping.org/wbox/, 2007.
• 24: X Wang and M Chen. Cluster-level Feedback Power Control for Performance Optimization. 14th IEEE International Symposium on High-Performance Computer Architecture (HPCA 2008), February 2008.
• 25: Zhikui Wang, Xiaoyun Zhu, Cliff McCarthy, Partha Ranganathan, and Vanish Talwar. Feedback Control Algorithms for Power Management of Servers. In Third International Workshop on Feedback Control Implementation and Design in Computing Systems and Networks (FeBid), Annapolis, MD, June 2008.

INDEX TERMS

Primary Classification:
  C. Computer Systems Organization
  C.4 PERFORMANCE OF SYSTEMS
      Subjects: Modeling techniques

General Terms:
Experimentation, Measurement, Performance, Theory

Keywords:
data center, power management, power-to-frequency, response time, server farm