A CASE STUDY ON: THE EFFECTS OF HIGHER OPERATING TEMPERATURES ON DATA CENTER EFFICIENCY Victor Garcia Director of Facilities Brocade October 24, 2012 Jeff Stein, PE Principal Taylor Engineering 2010 Brocade Communications Systems, Inc. Proprietary Information 1
Agenda Background Hypothesis Experiment Findings Next Steps 2
Brocade San Jose Labs 75,000 square feet of labs 1,900 racks 3 kw/rack 6 MW Total Building Load Hot aisle containment In-row cooling Water side economizer PUE of 1.20 3
Hypothesis Lowering the cold aisle temperature setpoint reduces IT power but the HVAC penalty is bigger than the IT savings IT fan power goes down But HVAC fan and chiller plant power goes up more 4
IRC Control Sequence 95% 75% Fan Speed Max SAT = 80F 50% Supply Air Temp Setpoint Min SAT = 66F 0% 50% 75% 100% IRC Group Reset Point 5
Temperature Sensors Before On rack In front of equipment After Hanging in cold row Uniform temperatures 6
Cold Aisle Setpoint Experiment Raise and lower cold aisle temperature setpoint in several cold aisles Monitor IRC fan power, IT gear power 7
Area of Experiments 3 rd Floor 12 rows 9 cold aisles 2 nd floor 6 rows 4 cold aisles 8
Row 2T 9
Row 2T (server/storage gear) 1.5% increase in IT KW per 5 degf 10
Row 2T 11
Row 2T 12
Row 2R,2S (mixed gear) 1.2% increase in KW per 5 deg F 13
Row 3C (IP gear) No relationship between kw and Max Temp 14
Row 3C, 3D (IP/Servers) 2% step change at 85F 15
Row 3E, 3F (mixed, storage) 4% step change at 84F 16
Summary of IT Power Changes Row Gear Increasing Cold Aisle Temp results in: 2R Mixed not valid due to IT changes 2R, 2S mixed 1.2% increase in IT per 5F increase in cold aisle setpt 2S, 2T mixed 1.5%/5F positive relationship 2T server/storage 1.5%/5F positive relationship 3A IP/Servers not valid due to IT changes 3A, 3B IP/Servers not valid due to IT changes 3B mixed positive but also some IT changes 3C IP no relationship 3C, 3D IP/Servers 2% step change at 85F 3D Servers no relationship 3E SAN/Server no relationship 3E, 3F mixed (E) and storage (F) 4% step change at 85F 3F, 3G servers 1.2%/5F positive relationship 17
8% avg decrease in fan speed for 5F increase in cold aisle setpoint 18
8% increase in fan speed is a 30-50% increase in fan power 19
Plant Power: 20% increase in plant power per 5F increase in wetbulb 5F increase in chilled water return temp. would have similar effect 20
Water economizer serves more load as supply and return water temperatures increase Water Entering Economizer Water Leaving Economizer Water Leaving Chiller 21
Water economizer serves more load as supply and return water temperatures increase Load served by Economizer Load served by Chillers 22
Economizer Operation Last 30 Days 23
Results: Lowering Setpoint Increases Total Power 24
Sensitivity Analysis Q: How big a decrease in IT power makes the IT savings greater than the HVAC penalty? A: IT power would have to decrease at least 8% per 5F decrease in cold aisle setpoint We saw only a 1.2% decrease per 5F A recent study by APC/Dell found a 3% decrease per 5F (www.apcmedia.com/salestools/jben-7ktr88_r1_en.pdf ) 25
Conclusions If the HVAC system has an economizer and has good part load performance (e.g. VFD fans) then the hotter the cold aisles the lower the total power consumption Commissioning is key Continuous monitoring Containment integrity Cold aisle sensor placement Incorrectly oriented gear Rogue zones 26
Next Steps More testing and data analysis Total building level experiment Monitor loads at equipment level Hotter cold aisles is more efficient but does it effect IT gear reliability? IT gear fan control algorithms More commissioning 27
Thank You 28