Exploring the Potential of CMP Core Count Management on Data Center Energy Savings Ozlem Bilgir * Margaret Martonosi * Qiang Wu * Princeton University.

Slides:



Advertisements
Similar presentations
George Nychis✝, Chris Fallin✝, Thomas Moscibroda★, Onur Mutlu✝
Advertisements

Demand Response in Ontario Paul Grod, CEO, Rodan Energy July 11, 2013.
Chapter 5: CPU Scheduling
Greening Backbone Networks Shutting Off Cables in Bundled Links Will Fisher, Martin Suchara, and Jennifer Rexford Princeton University.
Dynamic Power Redistribution in Failure-Prone CMPs Paula Petrica, Jonathan A. Winter * and David H. Albonesi Cornell University *Google, Inc.
Doc.: IEEE /037r1 Submission March 2001 Khaled Turki et. al,Texas InstrumentsSlide 1 Simulation Results for p-DCF, v-DCF and Legacy DCF Khaled.
Doc.: IEEE /0338r1 Submission March 2012 Hung-Yu Wei, National Taiwan UniversitySlide 1 DeepSleep: Power Saving Mode to Support a Large Number.
Challenge the future Delft University of Technology Overprovisioning for Performance Consistency in Grids Nezih Yigitbasi and Dick Epema Parallel.
1 A. Sshaikh, A. Greenberg; Nov 01 UCSC Sigcomm IMW Experience in Black-box OSPF Measurement Aman Shaikh, UCSC Albert Greenberg, AT&T Labs-Research.
/4/2010 Box and Whisker Plots Objective: Learn how to read and draw box and whisker plots Starter: Order these numbers.
Making a Line Plot Collect data and put in chronological order
1  1 =.
Year 6 mental test 15 second questions Numbers and number system Numbers and the number system, Measures and Shape.
Peer-to-peer and agent-based computing Basic Theory of Agency.
Predicting Performance Impact of DVFS for Realistic Memory Systems Rustam Miftakhutdinov Eiman Ebrahimi Yale N. Patt.
A Switch-Based Approach to Starvation in Data Centers Alex Shpiner and Isaac Keslassy Department of Electrical Engineering, Technion. Gabi Bracha, Eyal.
$100 $200 $300 $400 $100 $200 $300 $400 $100 $200 $300 $400 $100 $200 $300 $400 $100 $200 $300 $400.
Scheduling Algorithems
Feedback Control Real-Time Scheduling: Framework, Modeling, and Algorithms Chenyang Lu, John A. Stankovic, Gang Tao, Sang H. Son Presented by Josh Carl.
Power Aware Scheduling for AND/OR Graphs in Multi-Processor Real-Time Systems Dakai Zhu, Nevine AbouGhazaleh, Daniel Mossé and Rami Melhem PARTS Group.
Utility Optimization for Event-Driven Distributed Infrastructures Cristian Lumezanu University of Maryland, College Park Sumeer BholaMark Astley IBM T.J.
Sabyasachi Ghosh Mark Redekopp Murali Annavaram Ming-Hsieh Department of EE USC KnightShift: Enhancing Energy Efficiency by.
© 2007 IBM Corporation Steve Bowden Green Computing CTO IBM Systems & Technology Group UKISA Sustainable Computing in a Carbon Sensitive World Blue turning.
Improving DRAM Performance by Parallelizing Refreshes with Accesses
Managing Web server performance with AutoTune agents by Y. Diao, J. L. Hellerstein, S. Parekh, J. P. Bigu Jangwon Han Seongwon Park
Seungmi Choi PlanetLab - Overview, History, and Future Directions - Using PlanetLab for Network Research: Myths, Realities, and Best Practices.
A Survey of Web Cache Replacement Strategies Stefan Podlipnig, Laszlo Boszormenyl University Klagenfurt ACM Computing Surveys, December 2003 Presenter:
Asaf Cidon. , Tomer M. London
What is it and how do I know when I see it?
Energy-efficient Task Scheduling in Heterogeneous Environment 2013/10/25.
Zhou Peng, Zuo Decheng, Zhou Haiying Harbin Institute of Technology 1.
© 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice Better business outcomes equal better.
New England Developments in Demand Response and Smart Grid 2010 National Town Meeting on Demand Response and Smart Grid Henry Yoshimura, Director, Demand.
Area of triangles.
T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A 1 September 2005MC-SSL Simulation 1 Analysis of Scalable Security – MC-SSL Simulation Reducing.
Introduction to Queuing Theory
Test B, 100 Subtraction Facts
Chapter 15: Quantitatve Methods in Health Care Management Yasar A. Ozcan 1 Chapter 15. Simulation.
Partial Products. Category 1 1 x 3-digit problems.
University of Minnesota Optimizing MapReduce Provisioning in the Cloud Michael Cardosa, Aameek Singh†, Himabindu Pucha†, Abhishek Chandra
Modeling User Activities in a Large IPTV System Tongqing Qiu, Jun (Jim) Xu (Georgia Tech) Zihui Ge, Seungjoon Lee, Jia Wang, Qi Zhao (AT&T Lab – Research)
The Effects of Wide-Area Conditions on WWW Server Performance Erich Nahum, Marcel Rosu, Srini Seshan, Jussara Almeida IBM T.J. Watson Research Center,
Reducing Network Energy Consumption via Sleeping and Rate- Adaption Sergiu Nedevschi, Lucian Popa, Gianluca Iannaccone, Sylvia Ratnasamy, David Wetherall.
Exploring Energy-Latency Tradeoffs for Broadcasts in Energy-Saving Sensor Networks AUTHOR: MATTHEW J. MILLER CIGDEM SENGUL INDRANIL GUPTA PRESENTER: WENYU.
Dynamic Thread Assignment on Heterogeneous Multiprocessor Architectures Pree Thiengburanathum Advanced computer architecture Oct 24,
Thread Criticality Predictors for Dynamic Performance, Power, and Resource Management in Chip Multiprocessors Abhishek Bhattacharjee Margaret Martonosi.
© 2006, François Brouard Case Real Group François Brouard, DBA, CA January 6, 2006.
1 MemScale: Active Low-Power Modes for Main Memory Qingyuan Deng, David Meisner*, Luiz Ramos, Thomas F. Wenisch*, and Ricardo Bianchini Rutgers University.
A Cyber-Physical Systems Approach to Energy Management in Data Centers Presented by Chen He Adopted form the paper authors.
RELATED BACKGROUND WORK OZLEM BILGIR. OUTLINE 1- Gandhi et al., Optimal Power Allocation in Server Farms, SIGMETRICS’09 2-Chen et al., Managing Server.
Control System for Energy Efficient Data Centers Ozlem Bilgir.
System-Wide Energy Minimization for Real-Time Tasks: Lower Bound and Approximation Xiliang Zhong and Cheng-Zhong Xu Dept. of Electrical & Computer Engg.
ECE 510 Brendan Crowley Paper Review October 31, 2006.
Folklore Confirmed: Compiling for Speed = Compiling for Energy Tomofumi Yuki INRIA, Rennes Sanjay Rajopadhye Colorado State University 1.
PARAID: The Gear-Shifting Power-Aware RAID Charles Weddle, Mathew Oldham, An-I Andy Wang – Florida State University Peter Reiher – University of California,
OPTIMAL SERVER PROVISIONING AND FREQUENCY ADJUSTMENT IN SERVER CLUSTERS Presented by: Xinying Zheng 09/13/ XINYING ZHENG, YU CAI MICHIGAN TECHNOLOGICAL.
Baoxian Zhao Hakan Aydin Dakai Zhu Computer Science Department Computer Science Department George Mason University University of Texas at San Antonio DAC.
IBM Research © 2010 IBM Corporation Guarded Power Gating in a Multi-core Setting Niti Madan, Alper Buyuktosunoglu, Pradip Bose, IBM T.J.Watson June 2010.
Thread Criticality Predictors for Dynamic Performance, Power, and Resource Management in Chip Multiprocessors Abhishek Bhattacharjee and Margaret Martonosi.
Critical Power Slope Understanding the Runtime Effects of Frequency Scaling Akihiko Miyoshi, Charles Lefurgy, Eric Van Hensbergen Ram Rajamony Raj Rajkumar.
1 Queuing Analysis Overview What is queuing analysis? - to study how people behave in waiting in line so that we could provide a solution with minimizing.
Towards Dynamic Green-Sizing for Database Servers Mustafa Korkmaz, Alexey Karyakin, Martin Karsten, Kenneth Salem University of Waterloo.
BarrierWatch: Characterizing Multithreaded Workloads across and within Program-Defined Epochs Socrates Demetriades and Sangyeun Cho Computer Frontiers.
Ensieea Rizwani An energy-efficient management mechanism for large-scale server clusters By: Zhenghua Xue, Dong, Ma, Fan, Mei 1.
Energy Aware Network Operations
ElasticTree Michael Fruchtman.
Frequency Governors for Cloud Database OLTP Workloads
TimeTrader: Exploiting Latency Tail to Save Datacenter Energy for Online Search Balajee Vamanan, Hamza Bin Sohail, Jahangir Hasan, and T. N. Vijaykumar.
Reducing Total Network Power Consumption
Chih-Hsun Chou Daniel Wong Laxmi N. Bhuyan
Presentation transcript:

Exploring the Potential of CMP Core Count Management on Data Center Energy Savings Ozlem Bilgir * Margaret Martonosi * Qiang Wu * Princeton University Facebook, Inc.

Low Utilization Example: 1-week CPU utilization for Facebook servers CPU utilizations stay below 40% most of the time High Idle Power Idle power can be as much as 60% of max power Waste of power Data Center Inefficiencies CPU Utilization (%)

Previous Work on Data Center Power Management Server Consolidation [Pinheiro et al. 2003, Chen et al. 2008, etc.] Dynamic Voltage Frequency Scaling [Chen et al. 2005, Bertini et al. 2010, etc.] Power-gating in CMPs [Leverich et al. 2009, Madan et al. 2010, Kumar et al. 2009] Server reboots are not very desirable because of o Frequent code pushes, importance of robustness, high boot latencies, etc. DVFS leverage is decreasing because o Operating voltages are decreasing, prominence of leakage power Reducing core power with power-gating will be used in this work Abstraction layer between latency and core count

Our Envisioned System Front End Device N 1 N 2 N 3 N 4 Server 1Server 2Server 3Server 4 Assume 1 core in each server is ON tata Service times vary with exponential distribution tdtd Latency = t d – t a Latency = Queuing Time + Service Time Set of Research Questions: How many cores? Which cores? How are answers affected by boot time, bursts, etc.?

Our Approach: Multi-Server CMP Core Count Management Consolidate load to only a subset of cores in multi- server CMP systems Put the other cores in low power state Decide which cores to keep ON at a global level Constraint: Latency requirement is satisfied

Outline Motivation System and Research Overview Decision of ON Core Count Core Count Management Alternatives Performance and Power Models Methodology Results Conclusion

Decision of ON Core Count Look-up table based ON core count decision maker Gives total necessary number of ON cores for a given latency goal and an observed load rate ON core count of servers change at every period, T Created beforehand by observing obtained latency at different load rate and ON core count Core Count % Latency Goal 150msec 0% 4% 9% … 86% 200msec 1% 7% 13% … 97% 250msec 3% 9% 14% … 100% 10%

Core Count Management Alternatives Round-Robin Scheme Same Server Scheme Chip Turn On/Off Scheme Resource contention effect is low - Opportunity to dedicate empty servers to other applications Better performance under bursts

Power and Performance Models Power Performance Per-core service time is affected by contention C ratio is contention ratio

Outline Motivation System and Research Overview Decision of ON Core Count Core Count Management Alternatives Performance and Power Models Methodology Workload Parameters Results Conclusion

Real data from Facebook Stochastic Data Exponential dist. with stable means at 5%, 40% and 85% Workload Time

All Parameters ExplanationValue Server Count4 Core Count per Server4 Base Service Time100msec Service Time DistributionExponential Distribution Inter-arrival Time DistributionExponential Distribution Control Period10min Contention Ratio0%, 15% Idle Core Power40% 75% Latency Goal250msec, 150msec

Results: Effect of Load Rate on Energy Energy savings are greater at low load rates For 5% load, 80% savings For Facebook load, 35% savings At high load rates, most cores are ON, hence less savings SS and RR behave same because C ratio =0

Results: Effect of Load Rate on Latency All of them satisfy the latency goal Obtained latencies are much better than the latency goal With fewer cores, obtained latency would exceed latency goal

Results: Effect of Contention on Energy Energy consumption is affected by contention more in SS Increase in Same-Server is 30% Increase in Round-Robin is13%

Results: Effect of Latency Goal on Energy Tighter latency goal -> More cores ON At 150msec, SS energy consumption increase by 10% CTO energy consumption increases by 6%

Conclusion Using stochastic simulation and real workload, a range of core count management issues in CMPs explored Our CMP core count control mechanism connects high level information (latency) to low-level power management (core count) 35% core energy can be saved with Facebook workload

Exploring the Potential of CMP Core Count Management on Data Center Energy Savings Ozlem Bilgir * Margaret Martonosi * Qiang Wu * Princeton University Facebook, Inc.