SlideShare a Scribd company logo
1 of 9
Download to read offline
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Scott Pakin
Applied Computer Science Group
Los Alamos National Laboratory
14 November 2012
Slide 1
Cool Supercomputing:
Keepin’ it Real
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Questions from the Organizers
 What is good and bad about the current state of the art in tools and
techniques for optimizing power on large-scale systems?
 How much more needs to be done to make power a first-class citizen
for future extreme-scale systems?
Slide 2
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Current State of the Art
 Good
• Lots of work being done to manage power throughout the system
• Algorithms, compilers, job schedulers, operating systems, architecture
 Bad
• Most of this work is totally oblivious to reality
Slide 3
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
The Wrong Way to Think about Power
 Researcher: “If you use my {run-time
system, compiler, language, etc.}, your
application will draw x% less power y%
of the time and degrade performance
by only z%”
 Financial analyst (gov’t): “Our budgets
don’t carry over across fiscal years;
drawing x% less power doesn’t save us
any money”
 Facilities engineer: “We have to
allocate infrastructure for worst-case
usage; y < 100% is useless”
 User: “What?!? You’re degrading my
performance by z%. What did I ever do
to you?
Slide 4
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
The Disconnect
 Users and application developers don’t
care about power efficiency
• They don’t pay for power
• They barely know how to handle scalability,
let alone programming for power efficiency
• It’s not worth their time to restructure code
for power efficiency
• Preferred metric: ED∞P
 Race-to-halt does better than most
researchers give it credit for
• DRAM, power supplies, I/O devices,
various other components draw power
whether used or not
• Implication is that energy is minimized
when these are used for as little time as
possible
Slide 5
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Race to Halt is Hard to Beat
 Power data for xRAGE on a
150-node Sandy Bridge +
InfiniBand cluster
• 109 W/node idle vs.
332 W/node at max. perf.
 Best one can do
• Reduce power by 2/3
• Increase run time by <2/3 to
come out ahead energy-wise
• Possible? Doubtful
 Change of goals
• Reduce baseline power draw
• Get most performance for a
given power budget
Slide 6
0
50
100
150
200
250
300
350
Idle Max. perf.
Nodepower(W)
Other
Uncore
Core
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
The Right Way to Think about Power
Slide 7
0
2,000
4,000
6,000
8,000
10,000
12,000
14,000
16,000
18,000
0 2,000 4,000 6,000 8,000 10,000 12,000
Peakperformance(Top500Rmax
inTflops/s)
Power budget (kW)
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Making Power a First-Class Citizen
 Necessary pain at extreme scale
• Applications are granted a
maximum power draw for the
course of their execution
 Pain relief (naproxen)
• Give application developers the
mechanisms needed to stay within
their budget
• Libraries, language constructs, etc.
 Pain relief (homeopathic)
• Throttle performance if application
tries to exceed its power cap
• Bonus points: Coschedule high-
and low-power applications
• (You can go over budget if you find
a patsy who can stay under budget)
Slide 8
Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA
U N C L A S S I F I E D
Closing Thoughts
“Gotta give us what we want.
Gotta give us what we need.
…
To revolutionize make a change,
Nothin's strange
…
What we need is awareness;
we can’t get careless.
…
Lemme hear you say,
Fight the power.”
— Public Enemy,
Fight the Power
Slide 9

More Related Content

Viewers also liked

Herramientas
HerramientasHerramientas
Herramientascarmenr25
 
Herramientas
HerramientasHerramientas
Herramientascarmenr25
 
Herramientas
HerramientasHerramientas
Herramientascarmenr25
 
Trends in Sensors, Wearable Devices and IoT
Trends in Sensors, Wearable Devices and IoTTrends in Sensors, Wearable Devices and IoT
Trends in Sensors, Wearable Devices and IoTWalt Maclay
 
Herramientas
HerramientasHerramientas
Herramientascarmenr25
 
Herramientas
HerramientasHerramientas
Herramientascarmenr25
 
Herramientas
HerramientasHerramientas
Herramientascarmenr25
 

Viewers also liked (7)

Herramientas
HerramientasHerramientas
Herramientas
 
Herramientas
HerramientasHerramientas
Herramientas
 
Herramientas
HerramientasHerramientas
Herramientas
 
Trends in Sensors, Wearable Devices and IoT
Trends in Sensors, Wearable Devices and IoTTrends in Sensors, Wearable Devices and IoT
Trends in Sensors, Wearable Devices and IoT
 
Herramientas
HerramientasHerramientas
Herramientas
 
Herramientas
HerramientasHerramientas
Herramientas
 
Herramientas
HerramientasHerramientas
Herramientas
 

Similar to Cool Supercomputing: Keepin' it Real

SS15_PresentationFinal
SS15_PresentationFinalSS15_PresentationFinal
SS15_PresentationFinalJohn Determan
 
Redefining "Clean IT": Rejecting Incremental Improvement
Redefining "Clean IT": Rejecting Incremental ImprovementRedefining "Clean IT": Rejecting Incremental Improvement
Redefining "Clean IT": Rejecting Incremental ImprovementPeter Coffee
 
Accelerators at ORNL - Application Readiness, Early Science, and Industry Impact
Accelerators at ORNL - Application Readiness, Early Science, and Industry ImpactAccelerators at ORNL - Application Readiness, Early Science, and Industry Impact
Accelerators at ORNL - Application Readiness, Early Science, and Industry Impactinside-BigData.com
 
Matthew Egan End of Assignment Presentation 2nd Rotation
Matthew Egan End of Assignment Presentation 2nd RotationMatthew Egan End of Assignment Presentation 2nd Rotation
Matthew Egan End of Assignment Presentation 2nd RotationMatthew Egan
 
Energy Exchange 2015 Track 3 session 3: High Performance Buildings Operations
Energy Exchange 2015 Track 3 session 3: High Performance Buildings OperationsEnergy Exchange 2015 Track 3 session 3: High Performance Buildings Operations
Energy Exchange 2015 Track 3 session 3: High Performance Buildings OperationsShanti Pless
 
Career Profile R2
Career Profile R2Career Profile R2
Career Profile R2JeffBullen
 
Ips connect 2015 marc mueller stoffels
Ips connect 2015 marc mueller stoffelsIps connect 2015 marc mueller stoffels
Ips connect 2015 marc mueller stoffelsjames_hamilton
 
Focus On Energy It Presentation 2009 12 02
Focus On Energy It Presentation 2009 12 02Focus On Energy It Presentation 2009 12 02
Focus On Energy It Presentation 2009 12 02JeremyMathews
 
SiS Intelligent Energy Management Platt 2007
SiS Intelligent Energy Management Platt 2007SiS Intelligent Energy Management Platt 2007
SiS Intelligent Energy Management Platt 2007guestb40d60
 
Modeling distribution networks with neplan
Modeling distribution networks with neplanModeling distribution networks with neplan
Modeling distribution networks with neplanYusuf A. KHALIL
 
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICES
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICESLEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICES
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICESMichael P Totten
 
How to Lower Android Power Consumption Without Affecting Performance
How to Lower Android Power Consumption Without Affecting PerformanceHow to Lower Android Power Consumption Without Affecting Performance
How to Lower Android Power Consumption Without Affecting Performancerickschwar
 
Power Presentation West Palm
Power Presentation West PalmPower Presentation West Palm
Power Presentation West Palmwbbates
 
1293702-1578722-diegoramos-1.pptx
1293702-1578722-diegoramos-1.pptx1293702-1578722-diegoramos-1.pptx
1293702-1578722-diegoramos-1.pptxssusere27980
 
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...Sayonsom Chanda
 
PosterPresentation
PosterPresentationPosterPresentation
PosterPresentationRaj Shekhar
 

Similar to Cool Supercomputing: Keepin' it Real (20)

SS15_PresentationFinal
SS15_PresentationFinalSS15_PresentationFinal
SS15_PresentationFinal
 
IGBC2014 - Stephen Selkowitz, Lawrence Berkeley National Laboratory
IGBC2014 - Stephen Selkowitz, Lawrence Berkeley National LaboratoryIGBC2014 - Stephen Selkowitz, Lawrence Berkeley National Laboratory
IGBC2014 - Stephen Selkowitz, Lawrence Berkeley National Laboratory
 
Redefining "Clean IT": Rejecting Incremental Improvement
Redefining "Clean IT": Rejecting Incremental ImprovementRedefining "Clean IT": Rejecting Incremental Improvement
Redefining "Clean IT": Rejecting Incremental Improvement
 
Accelerators at ORNL - Application Readiness, Early Science, and Industry Impact
Accelerators at ORNL - Application Readiness, Early Science, and Industry ImpactAccelerators at ORNL - Application Readiness, Early Science, and Industry Impact
Accelerators at ORNL - Application Readiness, Early Science, and Industry Impact
 
GloSho'14: Company Showcase - Maxout Renewables
GloSho'14: Company Showcase - Maxout Renewables GloSho'14: Company Showcase - Maxout Renewables
GloSho'14: Company Showcase - Maxout Renewables
 
Matthew Egan End of Assignment Presentation 2nd Rotation
Matthew Egan End of Assignment Presentation 2nd RotationMatthew Egan End of Assignment Presentation 2nd Rotation
Matthew Egan End of Assignment Presentation 2nd Rotation
 
Brochure es 2018
Brochure es 2018Brochure es 2018
Brochure es 2018
 
Energy Exchange 2015 Track 3 session 3: High Performance Buildings Operations
Energy Exchange 2015 Track 3 session 3: High Performance Buildings OperationsEnergy Exchange 2015 Track 3 session 3: High Performance Buildings Operations
Energy Exchange 2015 Track 3 session 3: High Performance Buildings Operations
 
Career Profile R2
Career Profile R2Career Profile R2
Career Profile R2
 
Ips connect 2015 marc mueller stoffels
Ips connect 2015 marc mueller stoffelsIps connect 2015 marc mueller stoffels
Ips connect 2015 marc mueller stoffels
 
Focus On Energy It Presentation 2009 12 02
Focus On Energy It Presentation 2009 12 02Focus On Energy It Presentation 2009 12 02
Focus On Energy It Presentation 2009 12 02
 
4.4_Micro Grid Design_Bello_EPRI/SNL Microgrid
4.4_Micro Grid Design_Bello_EPRI/SNL Microgrid4.4_Micro Grid Design_Bello_EPRI/SNL Microgrid
4.4_Micro Grid Design_Bello_EPRI/SNL Microgrid
 
SiS Intelligent Energy Management Platt 2007
SiS Intelligent Energy Management Platt 2007SiS Intelligent Energy Management Platt 2007
SiS Intelligent Energy Management Platt 2007
 
Modeling distribution networks with neplan
Modeling distribution networks with neplanModeling distribution networks with neplan
Modeling distribution networks with neplan
 
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICES
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICESLEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICES
LEAST-COST-&-RISK LIFECYCLE DELIVERED ENERGY SERVICES
 
How to Lower Android Power Consumption Without Affecting Performance
How to Lower Android Power Consumption Without Affecting PerformanceHow to Lower Android Power Consumption Without Affecting Performance
How to Lower Android Power Consumption Without Affecting Performance
 
Power Presentation West Palm
Power Presentation West PalmPower Presentation West Palm
Power Presentation West Palm
 
1293702-1578722-diegoramos-1.pptx
1293702-1578722-diegoramos-1.pptx1293702-1578722-diegoramos-1.pptx
1293702-1578722-diegoramos-1.pptx
 
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...
How to leverage Quantum Computing and Generative AI for Clean Energy Transiti...
 
PosterPresentation
PosterPresentationPosterPresentation
PosterPresentation
 

Recently uploaded

[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdfhans926745
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...Neo4j
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationRadu Cotescu
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Enterprise Knowledge
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Servicegiselly40
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxKatpro Technologies
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountPuma Security, LLC
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Scriptwesley chun
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)wesley chun
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processorsdebabhi2
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUK Journal
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityPrincipled Technologies
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationSafe Software
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessPixlogix Infotech
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024The Digital Insurer
 

Recently uploaded (20)

[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...Workshop - Best of Both Worlds_ Combine  KG and Vector search for  enhanced R...
Workshop - Best of Both Worlds_ Combine KG and Vector search for enhanced R...
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Scaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organizationScaling API-first – The story of a global engineering organization
Scaling API-first – The story of a global engineering organization
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...Driving Behavioral Change for Information Management through Data-Driven Gree...
Driving Behavioral Change for Information Management through Data-Driven Gree...
 
Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024Tata AIG General Insurance Company - Insurer Innovation Award 2024
Tata AIG General Insurance Company - Insurer Innovation Award 2024
 
CNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of ServiceCNv6 Instructor Chapter 6 Quality of Service
CNv6 Instructor Chapter 6 Quality of Service
 
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptxFactors to Consider When Choosing Accounts Payable Services Providers.pptx
Factors to Consider When Choosing Accounts Payable Services Providers.pptx
 
Breaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path MountBreaking the Kubernetes Kill Chain: Host Path Mount
Breaking the Kubernetes Kill Chain: Host Path Mount
 
Automating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps ScriptAutomating Google Workspace (GWS) & more with Apps Script
Automating Google Workspace (GWS) & more with Apps Script
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
Exploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone ProcessorsExploring the Future Potential of AI-Enabled Smartphone Processors
Exploring the Future Potential of AI-Enabled Smartphone Processors
 
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdfUnderstanding Discord NSFW Servers A Guide for Responsible Users.pdf
Understanding Discord NSFW Servers A Guide for Responsible Users.pdf
 
Boost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivityBoost PC performance: How more available memory can improve productivity
Boost PC performance: How more available memory can improve productivity
 
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time AutomationFrom Event to Action: Accelerate Your Decision Making with Real-Time Automation
From Event to Action: Accelerate Your Decision Making with Real-Time Automation
 
Advantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your BusinessAdvantages of Hiring UIUX Design Service Providers for Your Business
Advantages of Hiring UIUX Design Service Providers for Your Business
 
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
Bajaj Allianz Life Insurance Company - Insurer Innovation Award 2024
 

Cool Supercomputing: Keepin' it Real

  • 1. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Scott Pakin Applied Computer Science Group Los Alamos National Laboratory 14 November 2012 Slide 1 Cool Supercomputing: Keepin’ it Real
  • 2. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Questions from the Organizers  What is good and bad about the current state of the art in tools and techniques for optimizing power on large-scale systems?  How much more needs to be done to make power a first-class citizen for future extreme-scale systems? Slide 2
  • 3. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Current State of the Art  Good • Lots of work being done to manage power throughout the system • Algorithms, compilers, job schedulers, operating systems, architecture  Bad • Most of this work is totally oblivious to reality Slide 3
  • 4. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D The Wrong Way to Think about Power  Researcher: “If you use my {run-time system, compiler, language, etc.}, your application will draw x% less power y% of the time and degrade performance by only z%”  Financial analyst (gov’t): “Our budgets don’t carry over across fiscal years; drawing x% less power doesn’t save us any money”  Facilities engineer: “We have to allocate infrastructure for worst-case usage; y < 100% is useless”  User: “What?!? You’re degrading my performance by z%. What did I ever do to you? Slide 4
  • 5. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D The Disconnect  Users and application developers don’t care about power efficiency • They don’t pay for power • They barely know how to handle scalability, let alone programming for power efficiency • It’s not worth their time to restructure code for power efficiency • Preferred metric: ED∞P  Race-to-halt does better than most researchers give it credit for • DRAM, power supplies, I/O devices, various other components draw power whether used or not • Implication is that energy is minimized when these are used for as little time as possible Slide 5
  • 6. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Race to Halt is Hard to Beat  Power data for xRAGE on a 150-node Sandy Bridge + InfiniBand cluster • 109 W/node idle vs. 332 W/node at max. perf.  Best one can do • Reduce power by 2/3 • Increase run time by <2/3 to come out ahead energy-wise • Possible? Doubtful  Change of goals • Reduce baseline power draw • Get most performance for a given power budget Slide 6 0 50 100 150 200 250 300 350 Idle Max. perf. Nodepower(W) Other Uncore Core
  • 7. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D The Right Way to Think about Power Slide 7 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 18,000 0 2,000 4,000 6,000 8,000 10,000 12,000 Peakperformance(Top500Rmax inTflops/s) Power budget (kW)
  • 8. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Making Power a First-Class Citizen  Necessary pain at extreme scale • Applications are granted a maximum power draw for the course of their execution  Pain relief (naproxen) • Give application developers the mechanisms needed to stay within their budget • Libraries, language constructs, etc.  Pain relief (homeopathic) • Throttle performance if application tries to exceed its power cap • Bonus points: Coschedule high- and low-power applications • (You can go over budget if you find a patsy who can stay under budget) Slide 8
  • 9. Operated by Los Alamos National Security, LLC for the U.S. Department of Energy’s NNSA U N C L A S S I F I E D Closing Thoughts “Gotta give us what we want. Gotta give us what we need. … To revolutionize make a change, Nothin's strange … What we need is awareness; we can’t get careless. … Lemme hear you say, Fight the power.” — Public Enemy, Fight the Power Slide 9