Energy Efficient High-tech Buildings

Transcription

1 Energy Efficient High-tech Buildings Can anything be done to improve Data Center and Cleanroom energy efficiency? Bill Tschudi Lawrence Berkeley National Laboratory

2 Acknowledgements California Energy Commission Pacific Gas and Electric Company Federal Energy Management Program (FEMP) 7 x 24 Exchange Uptime Institute Rocky Mountain Institute NYSERDA E Source Rumsey Engineers EYP Mission Critical Facilities Industry Partners (Too many to name all)

3 Data Center efficiency Benchmarking Case study example Power conversions Selected opportunities for improvement

4 We also operate data centers National Energy Research Scientific Computing supercomputer center in Oakland

5 California energy research related to data centers Energy research roadmap Case studies and energy benchmarking Best practice identification Self benchmarking protocol Investigate efficiency of power supplies in IT equipment Investigate efficiency of UPS systems Metrics for computing performance vs. energy Technology transfer Demonstration projects

6 Data center efficiency opportunity Industry professionals, LBNL and others brainstormed efficiency improvements at RMI Charrette (2003) Practical (near term) solutions as well as longer term concepts were identified Available through RMI website:

7 Data center energy roadmap Input through workshops, conferences, and contacts with industry professionals. Participation in design charrette facilitated by the Rocky Mountain Research Institute (RMI) Selected research areas are being pursued Available through LBNL website: admapfinal.pdf

8 Why benchmark data centers? Utility load growth planning Baseline energy use System and component efficiency comparisons Best practices using current technology Identify areas where further work is required

9 April 10, 2003 San Jose Mercury News A new power plant is up and running in San Jose's Alviso neighborhood, but the massive Internet server farm that it was supposed to fuel is nowhere in sight. The Los Esteros Critical Energy Facility, a 180-megawatt plant built by Calpine in North San Jose, was designed to power an adjacent Internet server farm by U.S. Dataport. The server farm never broke ground -- and company officials didn't return calls Wednesday to say if or when it might -- but Calpine proceeded with the plant anyway, after securing a three-year deal with the state Department of Water Resources to buy power. Company and state officials say the plant is still needed, even though the state's infamous energy crunch of is long over. 180 MW: 900,000 sq.ft. x 200 W/sq.ft

10 Case studies/benchmarks California Storage device and router Mfgs. Banks Web hosting facilities Internet service provider State tax center Federal facilities New York Recovery center (hosting) Financial institution

11 Electricity flow in Data Centers HVAC system lights, office space, etc. uninterruptible power local distribution lines to the building, 480 V UPS PDU computer racks computer equipment backup diesel generators UPS = Uninterruptible Power Supply PDU = Power Distribution Unit;

12 Metrics IT equipment load intensity: W/sq. ft. (electrically active space Uptime definition) UPS losses: % Chilled water: kw/ton; W/sq.ft. End use pie chart: W/end use; W/sq.ft. Occupancy: % full (subjective) % of design load (readout from UPS or PDU)

13 2003 IT equipment loads (W/Sq.Ft. of electrically active floor space) Computer Load Density W/sq.ft Facility Average 27 +/-

14 Projecting fully loaded conditions (W/Sq.Ft. of electrically active floor space) Current and Projected Load Intensity W/sq. ft Facility Projected Average 44 Current Computer Load Projected Computer Load

15 Distribution of computer room power reported to Uptime Institute Fraction of total floor area in sample Source: Uptime Institute, Number of facilities Computer room power density W/square foot Total floor area Million square feet Computer room UPS power (Watts/square foot)

16 2005 IT equipment benchmarks IT equipment load Average 52 w/sf LBNL NERSC supercomputer Watts/sf Data center identifier

17 End-use pie charts vary Office Space Conditioning 1% Lighting 2% Other 13% Computer Loads 67% Electrical Room Cooling 4% Cooling Tower Plant 4% Data Center Server Load 51% HVAC - Air Movement 7% Data Center CRAC Units 25% Lighting 2% HVAC - Chiller and Pumps 24%

18 A better ratio is when infrastructure loads account for a smaller percentage Computer Loads 67% HVAC - Air Movement 7% Lighting 2% HVAC - Chiller and Pumps 24%

19 Effectiveness of HVAC systems HVAC (as a % of total load) 60% 50% % of total load 40% 30% 20% 10% 0% Data Center Identifier

20 Index of performance The Uptime Institute proposed a metric to evaluate the total efficiency of infrastructure systems: Index of performance = building systems KW UPS output (i.e. ratio of building systems to IT equipment load)

21 Total chilled water system efficiency Total Average Efficiency kw/ton kw/ton Lower is better Average 1.69 Facility

22 Chiller comparison 1.2 Average kw / ton Lower is better Facility

23 Chilled water systems efficiencies 1.8 kw / ton (lower is better) Cooling Tower CW Pumps CHW Pumps Chiller Water Cooled 42F Air Cooled 42F Air Cooled 40F Air Cooled 48F Air Cooled 50F Water Cooled 40F Water Cooled 38F Water Cooled 36F Water Cooled 44F Water Cooled 43F Fac. A Fac. B.1 Fac. B.2 Fac. B.2 Fac. B.2 Fac. C Fac. D Fac. E.1 Fac. E.2 Fac. F

24 Power loss in UPS systems Loss in UPS as a percent of total Average 7.7 % 10 Percent Facility Cooling loads compound this

25 How many times do data centers convert AC and DC? 5V Internal Drive In Battery/Charger Rectifier Bypass Inverter Out AC/DC PWM/PFC Switcher Unregulated DC To Multi Output Regulated DC Voltages 12V 3.3V 12V 1.5/2. DC/DC 5V 12V 3.3V DC/DC 1.1V- 1.85V External Drive I/O Memory Controller µ Processor SDRAM 3.3V Graphics Controller Voltage Regulator Modules AC/DC Multi output PS

26 Measured UPS efficiency Efficiency (%) Load Factor (%)

27 UPS factory measurements 100% Factory Measurements of UPS Efficiency (tested using linear loads) 95% Efficiency 90% 85% 80% Flywheel UPS 75% 70% Double-Conversion UPS Line-Interactive, Delta Conversion UPS 0% 20% 40% 60% 80% 100% Percent of Rated Active Power Load

28 Measured UPS losses 100% Factory Measurements of UPS Efficiency (tested using non-linear loads) 95% Efficiency 90% 85% 80% 75% 70% Double Conversion UPS Line-Interactive, Delta Conversion UPS 0% 20% 40% 60% 80% 100% Percent of Rated Active Power Load

29 Double conversion UPS systems can be more efficient today

30 Analyzing UPS performance in high efficiency option High Efficiency Mode 30% sag 10 cycle Double Conversion Mode 30% sag 30 cycle Input Voltage Output Voltage Input Voltage Output Voltage Volts Time (ms) Volts Time (ms) Source: EPRI Solutions In high efficiency mode, there is typically one cycle (16.6 msec for 60 Hz) of UPS output voltage deviation. Power supplies downstream of UPS can ride through this!

31 AC DC Losses Electricity use in a server DC/DC Losses Fans 72 Drives 41 PCI Cards 86 Processors Based on a typical dual processor 450W 2U Server; Approximately 160W out of 450W (35%) are losses in the power conversion process (Source: Brian Griffith: INTEL) Memory Chipset

32 Power supply opportunity HVAC Chilled Water Standby Generator Efficiency of Power Conversion Process IT Load (kw) UPS Losses (kw) Total Savings (kw) Lighting and Plug Loads UPS Losses Computing Load 65% % % % % % HVAC Fan Load Based on one case study approximately 4335 KW of a total of 8500 kw was IT load. Assuming a 65% existing baseline efficiency, the savings opportunity using 90% efficient conversion process is approximately 1300kW not including any savings from HVAC

33 Power supply efficiency today Full Load Efficiency > 68% Redundant System of Power Supplies for Servers 68

34 Measured power supply efficiency 85% 80% 75% % Efficiency 70% 65% 60% 55% 50% Average of All Servers 45% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % of Nameplate Power Output

35 Standby generation loss Several load sources Heaters Battery chargers Transfer switches Fuel management systems Heaters (many operating hours) use more electricity than the generator will ever produce (few operating hours) Opportunity may be to reduce or eliminate heating, batteries, and chargers

36 Standby generator heater

37 Additional benchmarks Computations per Watt Ability to enter sleep mode Nameplate vs. Actual Comparisons Others? IT Equipment UPS Chillers Transformers

38 California case studies LBNL sub-contractors, Rumsey Engineers and EYP Mission Critical Facilities, performed site data collection and preliminary analysis, provided efficiency recommendations, and case studies reports.

39 Case study example: Facility 8 % of total load 60% 50% 40% 30% 20% 10% HVAC (as a % of total load) The worst and the best ratios were both in facility 8 0% Data Center Identifier

40 Facility 8 site characteristics Data Center ,200 sq ft 6 UPS s 3 per side Redundancy: n+1 at PDU level, n+2 at UPS level Overhead ducted air distribution Air-cooled constant volume CRAC units

41 Facility 8 site characteristics Data Center ,000 sq ft 5 UPS s Redundancy: n+1 at PDU level Overhead ducted air distribution Central chilled water plant Central air handling system Variable speed chiller, secondary pumps, air handlers

42 Facility 8 electricity end-use Data Center 8.1 Data Center 8.2 Total Power = 580 kw UPS Losses 6% Computer Loads 63% Total Power = 1700 kw Computer Loads 38% HVAC 54% UPS Losses 13% Lighting 2% Lighting 1% HVAC - Chilled Water Plant 14% HVAC - Air Movement 9%

43 How did they do it? Data Center 8.1 Air cooled CRAC units No economizing Constant Speed Fans Humidification control All CRAC units on Data Center 8.2 Optimal central chilled water plant Optimal central air handling units Little humidity control Good control Data Monitoring Gateways, EMCS

44 How could they have done even better? Data center 8.1 observations Disable humidification control Turn off CRAC units Shut off (rotate) surplus UPSs i.e. go from N+2 to desired N+1 Space temperature setpoints

45 How could they have done even better? Data center 8.2 observations Monitoring - chiller, total chiller plant kw/ton Run Cooling towers in parallel, nozzle replacement Chilled water setpoint Condenser water temperature reset

46 Energy efficiency opportunity Air management hot aisle-cold aisle, bypassing/short circuiting, underfloor congestion, high ceilings and adequate underfloor areas, use of modeling Air and water side economizers many hours of low cost cooling in CA UPS and power supply efficiencies, loading, redundancy strategies Temperature and humidity control eliminate CRAC unit fighting, ASHRAE thermal guidelines Chilled water plant optimization - efficient chillers, primary only pumping, etc. Variable speed drives pumps, chillers, fans Control strategies - setpoints, cooling tower staging Lighting controls

47 LBNL high-tech buildings website:

48 Thank you Questions?