Innovative Data Center Efficiency Techniques. Michael K Patterson, PhD PE Eco-technology Program Office Intel Corporation

Size: px

Start display at page:

Download "Innovative Data Center Efficiency Techniques. Michael K Patterson, PhD PE Eco-technology Program Office Intel Corporation"

Madeline McLaughlin
5 years ago
Views:

1 Innovative Data Center Efficiency Techniques Michael K Patterson, PhD PE Eco-technology Program Office Intel Corporation 1

2 Alternate titles.. Crisis or opportunity? Never build a new data center! Improve your PUE & your efficiency The hot aisle is supposed to be hot It s all about the denominator! 2

3 A Data Center Crisis or Opportunity? Projected Data Center Energy Use Under Five Scenarios Billions (kwh / year) % of total US electricity usage 1.5% of total US. electricity usage 2.9% of projected total U.S. electricity use Historical Trends Current Efficiency Trends Improved Operation Best Practice forecast State-ofthe-Art Where do you want to be? Source: EPA Report to Congress on Server and Data Center Energy Efficiency; August 2,

4 Moore s Law Alive and Well in the Data Center Where would YOU rather be? TFlops 25 racks 512 Servers 90 m2 128 kw TFlops 1 rack 53 Servers 4 m2 21 kw 4

But what does this mean to our efficiency challenges? As we have commoditized the server, we have un-commoditized the data center. This is NOT a crisis, instead it s an opportunity to differentiate.

5 But what does this mean to our efficiency challenges? As we have commoditized the server, we have un-commoditized the data center. This is NOT a crisis, instead it s an opportunity to differentiate. Those who apply the right planning and engineering will prosper, those who do not will suffer. Good engineering will allow higher efficiency, lower TCO, and more computing capability TFlops 1 rack 64 Servers 40 sq ft 30 kw 5

6 The real world 6 6

7 Alternate titles.. Crisis or opportunity? Never build a new data center! Improve your PUE & your efficiency The hot aisle is supposed to be hot It s all about the denominator! 7

Server Refresh Example in 2010 2005 2010 Efficiency Refresh 15:1

new quad core Servers 15 Racks of Single Core Servers OR

Performance 8% Annual Energy Costs Reduction (estimated) Source:

Performance comparison using SPECjbb2005 bops (business

Results have been estimated based on internal Intel analysis and

8 Server Refresh Example in Efficiency Refresh 15:1 More than 90% Annual Energy Cost Reduction (estimated) 1 rack of new quad core Servers 15 Racks of Single Core Servers OR Performance Refresh 1:1 15 racks of quad core Servers Up to 15x Performance 8% Annual Energy Costs Reduction (estimated) Source: Intel estimates as of Jan Performance comparison using SPECjbb2005 bops (business operations per second). Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance. 8

9 Alternate titles.. Crisis or opportunity? Never build a new data center! Improve your PUE & your efficiency The hot aisle is supposed to be hot It s all about the denominator! 9

10 10 Source: The Green Grid

11 Infrastructure Optimization Effects DC 2020 Part I. Power [kw] Status quo CMO basic CMO step 1 CMO step 2 5kW/Rack CMO step 3 CMO step 4 CMO best case ΔPUE ~ 20% FMO ASHRAE FMO HW limit Total Load 1,9 IT Load 1,8 PUE 1,7 1,6 1,5 1,4 1,3 1,2 1,1 1 PUE 1. Reduce Leakage 2. Optimize Fan Speed 3. Apply Doors, opt. Fan 4. Grating Tiles, optimize Fan 5. Complete enclosure, opt. Fan 6. EC Fan, Water supply 14 C 7. Water supply 24 C, Server Inlet 27 C 8. Water supply 34 C, Server Inlet 35 C 11

12 ERE Development PUE = PUE = Total Energy IT Energy Cooling + Power + Lighting + IT IT ERE = ERE = Total Energy - Reused Energy IT Energy Cooling + Power + Lighting + IT - Reused IT 12

13 Data Center Boundary Consideration and ERF Cooling E Rejected or re-used?? A Utility IT F B UPS C PDU D PUE = A + B D ERE = A + B - F D Data Center Boundary 13

14 Proper use of ERE Non-DC space rejected H Utility A Cool E ICT G Absorp Chiller F I B 14 PUE = UPS A + B D C PDU ERE = A + B - F D Both PUE & ERF valid metrics D Data Center Boundary

15 Improper use of ERE rejected H Utility A Cool E ICT G Absorp Chiller F I B 15 PUE = UPS A + B D C PDU ERE = A + B - F D PUE is the right metric, ERE is not D Data Center Boundary

16 PUE, ERE, and ERF Ranges 1 PUE PUE = Cooling + Power + Lighting + IT IT 0 ERE ERE = Cool + Pwr + Light IT + IT - Reused 0 ERF 1 ERF = Reuse Energy Total Energy ERE = (1- ERF) PUE 16

17 Alternate titles.. Crisis or opportunity? Never build a new data center! Improve your PUE & your efficiency The hot aisle is supposed to be hot It s all about the denominator! 17

18 Scientific Research Computing Facility Stanford University Cooling Load Factor = 6% 18 Graphic courtesy of Stanford University

19 19 CERN cold aisle containment

20 Alternate titles.. Crisis or opportunity? Never build a new data center! Improve your PUE & your efficiency The hot aisle is supposed to be hot It s all about the denominator! 20

21 PUEs: Reported and Calculated PUE EPA Energy Star Average 1.91 Intel 1.41 T-Systems & Intel DC2020 Test Lab 1.24 Google 1.16 National Center for Atmospheric Research (NCAR) 1.10 Yahoo 1.08 National Renewable Energy Laboratory (NREL) 1.06 PUE Total Facility Energy IT Energy Cooling Power IT Misc IT 21

22 Xeon Platform Efficiency Trends Performance (k_ops) Utilization (Watts) Efficiency (Ops/Watt) Higher Performance AT LOWER POWER Performance and power consumption results are based on certain tests measured on specific computer systems. Any difference in system hardware, software or configuration will affect actual performance. Configurations: Two-socket Systems, Test Results for SPECpower_ssj2008, Testing by Hewlett-Packard. For more information go to 22

23 23

Server Efficiency Focus Global Impact Power Demand (GWatt) 12 10 8 6 Volume Server Global Power Demand Trend Estimate 2006 Up to 40% Reduction Trend Estimate 2010

5X Compute Capacity Grows 9X Total Server Energy Consumption Stays Constant The Tide is Turning on Data Center Energy Consumption 24 Source: Intel Estimates May

measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance.

24 Server Efficiency Focus Global Impact Power Demand (GWatt) Volume Server Global Power Demand Trend Estimate 2006 Up to 40% Reduction Trend Estimate Year By 2016 Number of Servers Increases 1.5X Compute Capacity Grows 9X Total Server Energy Consumption Stays Constant The Tide is Turning on Data Center Energy Consumption 24 Source: Intel Estimates May 2010 Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate performance of Intel products as measured by those tests. Any difference in system hardware or software design or configuration may affect actual performance. Buyers should consult other sources of information to evaluate the performance of systems or components they are considering purchasing. For more information on performance tests and on the performance of Intel products, Go to:

25 Moore s Law is the solution to Green Data Centers, not the problem.. Thank You! Questions? 25

Session 6: Data Center Energy Management Strategies

Session 6: Data Center Energy Management Strategies Data Center Energy Use and Opportunities: Applying Best Practices to meet Executive Order Requirements Dale Sartor, PE Lawrence Berkeley National Laboratory