ABSTRACT
Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Some microprocessors allow frequency and voltage scaling, which enables a system to reduce CPU performance and power when the CPU is not on the critical path. When properly directed, such dynamic frequency and voltage scaling can produce significant energy savings with little performance penalty.This paper presents an MPI runtime system that dynamically reduces CPU performance during communication phases in MPI programs. It dynamically identifies such phases and, without profiling or training, selects the CPU frequency in order to minimize energy-delay product. All analysis and subsequent frequency and voltage scaling is within MPI and so is entirely transparent to the application. This means that the large number of existing MPI programs, as well as new ones being developed, can use our system without modification. Results show that the average reduction in energy-delay product over the NAS benchmark suite is 10%---the average energy reduction is 12% while the average execution time increase is only 2.1%.
- N. D. Adiga et al. An overview of the BlueGene/L supercomputer. In Supercomputing, November 2002.]]Google ScholarDigital Library
- ASCI Purple Benchmark Suite. http://www.llnl.gov/asci/platiormstpurple/rfp/benchmarks/.]]Google Scholar
- D. Bailey, J. Barton, T. Lasinski, and H. Simon. The NAS parallel benchmarks. RNR-91-002, NASA Ames Research Center, August 1991.]]Google Scholar
- Ali Raza Butt, Chris Gniady, and Y. Charlie Hu. The performance impact of kernel prefetching on buffer cache replacement algorithms. In SIGMETRICS, pages 157--168, 2005.]]Google ScholarDigital Library
- K. W. Cameron, X. Feng, and R. Ge. Performance-constrained, distributed dvs scheduling for scientific applications on power-aware clusters. In Supercomputing, November 2005.]] Google ScholarDigital Library
- Enrique V. Carrera, Eduardo Pinheiro, and Ricardo Bianchini. Conserving disk energy in network servers. In Intl. Conference on Supercomputing, June 2003.]] Google ScholarDigital Library
- Jeffrey S. Chase, Darrell C. Anderson, Prachi N. Thakar, Amin Vahdat, and Ronald P. Doyle. Managing energy and server resources in hosting centres. In Symposium on Operating Systems Principles, 2001.]] Google ScholarDigital Library
- Guilin Chen, Konrad Malkowski, Mahmut Kandemir, and Padma Raghavan. Reducing power with performance contraints for parallel sparse applications. In Workshop on High-Performance, Power-Aware Computing, April 2005.]]Google Scholar
- A. Dhodapkar and J. Smith. Comparing phase detection techniques. In International Symposium on Microarchitecture, pages 217--227, December 2003.]] Google ScholarDigital Library
- Elmootazbellah Elnozahy, Michael Kistler, and Ramakrishnan Rajamony. Energy conservation policies for web servers. In Usenix Symposium on Internet Technologies and Systems, 2003.]] Google ScholarDigital Library
- E. N. (Mootaz) Elnozahy, Michael Kistler, and Ramakrishnan Rajamony. Energy-efficient server clusters. In Workshop on Mobile Computing Systems and Applications, Feb 2002.]]Google Scholar
- Mark E. Femal. Non-uniform power distribution in data centers for safely overprovisioning circuit capacity and booasting throughput. Master's thesis, North Carolina State University, Raleigh, NC, May 2005.]]Google Scholar
- Vincent W. Freeh, David K. Lowenthal, Feng Pan, and Nandani Kappiah. Using multiple energy gears in MPI programs on a power-scalable cluster. In Principles and Practices of Parallel Programming, June 2005.]] Google ScholarDigital Library
- Vincent W. Freeh, David K. Lowenthal, Rob Springer, Feng Pan, and Nandani Kappiah. Exploring the energy-time tradeoff in MPI programs on a power-scalable cluster. In International Parallel and Distributed Processing Symposium, April 2005.]] Google ScholarDigital Library
- Chris Gniady, Ali Raza Butt, and Y. Charlie Hu. Program-counter-based pattern classification in buffer caching. In OSDI, pages 395--408, 2004.]] Google ScholarDigital Library
- Chris Gniady, Y. Charlie Hu, and Yung-Hsiang Lu. Program counter based techniques for dynamic power management. In HPCA, pages 24--35, 2004.]] Google ScholarDigital Library
- Richard Goering. Current physical design tools come up short. EE Times, April 14 2000.]]Google Scholar
- Chung hsing Hsu and Wu chun Feng. A power-aware run-time system for high-performance computing. In Supercomputing, November 2005.]] Google ScholarDigital Library
- Chung-Hsing Hsu and Wu-chun Feng. Effective dynamic-voltage scaling through CPU-boundedness detection. In Fourth IEEE/ACM Workshop on Power-Aware Computing Systems, December 2004.]]Google Scholar
- M. Huang, J. Renau, and J. Torellas. Positional adaptation of processors: Application to energy reduction. In International Symposium on Computer Architecture, June 2003.]] Google ScholarDigital Library
- Nandani Kappiah, Vincent W. Freeh, and David K. Lowenthal. Just in time dynamic voltage scaling: Exploiting inter-node slack to save energy in MPI programs. In Supercomputing, November 2005.]] Google ScholarDigital Library
- Charles Lefurgy, Karthick Rajamani, Freeman Rawson, Wes Felter, Michael Kistler, and Tom W. Keller. Energy management for commerical servers. IEEE Computer, pages 39--48, December 2003.]] Google ScholarDigital Library
- Athanasios E. Papathanasiou and Michael L. Scott. Energy efficiency through burstiness. In Workshop on Mobile Computing Systems and Applications, October 2003.]]Google Scholar
- Eduardo Pinheiro, Ricardo Bianchini, Enrique V. Carrera, and Taliver Heath. Load balancing and unbalancing for power and performance in cluster-based systems. In Workshop on Compilers and Operating Systems for Low Power, September 2001.]]Google Scholar
- Rolf Rabenseifner. Automatic profiling of MPI applications with hardware performance counters. In PVM/MPI, pages 35--42, 1999.]] Google ScholarDigital Library
- Vivek Sharma, Arun Thomas, Tarek Abdelzaher, and Kevin Skadron. Power-aware QoS management in web servers. In IEEE Real-Time Systems Symposium, Cancun, Mexico, December 2003.]] Google ScholarDigital Library
- Timothy Sherwood, Erez Perelman, Greg Hamerly, and Brad Calder. Automatically characterizing large scale program behavior. In Architectural Support for Programming Languages and Operating Systems, October 2002.]] Google ScholarDigital Library
- Robert C. Springer IV, David K. Lowenthal, Barry Rountree, and Vincent W. Freeh. Minimizing execution time in MPI programs on an energy-constrained, power-scalable cluster. In ACM Symposium on Principles and Practice of Parallel Programming, March 2006.]] Google ScholarDigital Library
- M. Warren, E. Weigle, and W. Feng. High-density computing: A 240-node beowulf in one cubic meter. In Supercomputing, November 2002.]] Google ScholarDigital Library
- Qingbo Zhu, Francis M. David, Christo Devaraj, Zhenmin Li, Yuanyuan Zhou, and Pei Cao. Reducing energy consumption of disk storage using power-aware cache management. In High-Performance Computer Architecture, February 2004.]] Google ScholarDigital Library
Index Terms
- Adaptive, transparent frequency and voltage scaling of communication phases in MPI programs
Recommendations
Adaptive, transparent CPU scaling algorithms leveraging inter-node MPI communication regions
Although users of high-performance computing are most interested in raw performance, both energy and power consumption have become critical concerns. Because the CPU is often the major power consumer, some microprocessors allow frequency and voltage ...
Dynamic voltage and frequency scaling: the laws of diminishing returns
HotPower'10: Proceedings of the 2010 international conference on Power aware computing and systemsDynamic voltage and frequency scaling (DVFS) is a commonly-used power-management technique where the clock frequency of a processor is decreased to allow a corresponding reduction in the supply voltage. This reduces power consumption, which can lead to ...
Memory-aware energy-optimal frequency assignment for dynamic supply voltage scaling
ISLPED '04: Proceedings of the 2004 international symposium on Low power electronics and designDynamic supply voltage scaling (DVS) is one of the best ways to reduce the energy consumption of a device when there is a super-linear relationship between energy and supply voltage, and a pseudo-linear relationship between delay and supply voltage. ...
Comments