Sie sind auf Seite 1von 42

HVAC System Design

Taylor Engineering, LLC

Mark Hydeman, P.E., FASHRAE Taylor Engineering, LLC mhydeman@taylor-engineering.com

How do you effectively fight a fire?


it takes 2,000 to 3,000 times the volume of air to cool what you can with water! With air, or with water?

Taylor Engineering, LLC

State of the present: with air

Taylor Engineering, LLC

Air system design overview


Data center layout Airflow configurations

Distribution:

overhead or underfloor Control: constant or variable volume

Airflow issues Economizers Humidity control issues

Taylor Engineering, LLC 4

Data center layout


Server airflow front to back or front to back and top are recommended

Cold Aisle Hot Aisle

2004, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Thermal Guidelines for Data Processing Environments. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

Data center layout

Underfloor Supply

Cold Aisle Hot Aisle Only 1 pressure zone for UF!


2004, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Thermal Guidelines for Data Processing Environments. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

Data center layout


You can incorporate VAV on each branch
Overhead Supply

Cold Aisle Hot Aisle

2004, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Thermal Guidelines for Data Processing Environments. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

Typical temperature profile with UF supply


Too hot Too hot

Just right

Too cold Elevation at a cold aisle looking at racks


There are numerous references in ASHRAE. See for example V. Sorell et al; Comparison of Overhead and Underfloor Air Delivery Systems in a Data Center Environment Using CFD Modeling; ASHRAE Symposium Paper DE-05-11-5; 2005
Taylor Engineering, LLC 8

Typical temperature profile with OH supply


Too warm Too warm

Just right

Elevation at a cold aisle looking at racks

Taylor Engineering, LLC

Hot aisle lid

Aisle capping
End cap
Cold Aisle Caps
APC reprinted with permission

Cold Aisle Hot Aisle

2004, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Thermal Guidelines for Data Processing Environments. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

10

Aisle capping
LBNL has recently performed research on aisle capping
Cold Aisle Caps

Cold Aisle Hot Aisle

2004, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Thermal Guidelines for Data Processing Environments. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

11

Overhead (OH) vs. Underfloor (UF)


Issue Capacity Balancing Overhead (OH) Supply Limited by space and aisle velocity. Continuous on both outlet and branch. Underfloor (UF) Supply Limited by free area of floor tiles. Usually limited to incremental changes by diffuser type. Some tiles have balancing dampers. Also underfloor velocities can starve floor grilles! Only one pressure zone per floor, can provide multiple temperature zones. Commonly cold at bottom and hot at top. Generally worse. Worst. Hot or cold aisle possible.

Control Temperature Control First Cost Energy Cost Aisle Capping

Up to one pressure zone by branch. Most uniform. Best (if you eliminate the floor). Best. Hot or cold aisle possible.

Taylor Engineering, LLC

12

Airflow design disjoint


IT departments select servers and racks Engineers size the fans and cooling capacity Whats missing in this picture?

Taylor Engineering, LLC

13

Airflow with constant volume systems

Hot spots Higher hot aisle temperature Possible equipment failure or degradation

Taylor Engineering, LLC

V HVAC _ Supply V Servers

14

Airflow with constant volume systems

Taylor Engineering, LLC

V HVAC _ Supply V Servers

Least hot spots Higher air velocities Higher fan energy Reduced economizer effectiveness (due to lower return temperatures)

15

Airflow with constant volume systems


Note most of these observations apply to overhead and underfloor distribution With constant volume fans on the servers you can only be right at one condition of server loading! The solution is to employ variable speed server and distribution fans

Taylor Engineering, LLC 16

Airflow with variable volume systems

Partial flow condition

Best energy performance but difficult to control

Taylor Engineering, LLC

V HVAC _ Supply V Servers

17

How Do You Balance Airflow?

Spreadsheet CFD Monitoring/Site Measurements

Image from TileFlow http://www.inres.com/Products/TileFlow/tileflow.html, Used with permission from Innovative Research, Inc.

Taylor Engineering, LLC

18

Thermal report

From ASHRAEs Thermal Guidelines for Data Processing Environments Taylor Engineering, LLC 19

Whats the server airflow?


SUN V490 SUN V240 DELL 2850 DELL 6850 num fans 9 3 n/a n/a total CFM (max) 150 55.65 42 185 total CFM (min) 27 126 fan speed single speed variable 2 speed 2 speed fan control n/a inlet temp. 77F inlet 77F inlet Form Factor (in U's) 5 2 2 4 heat min config (btuh) 798 454 heat max config (btuh) 5,459 1,639 2,222 4,236 heat max (watts) 1,599 480 651 1,241 dT min config 13 3 dT max config 33 27 48 21 servers per rack 8 21 21 10 CFM/rack (hi inlet temp) 1,200 1,169 882 1,850 CFM/rack (low inlet temp) 1,200 567 1,260 max load / rack (kW) 13 10 14 12

Taylor Engineering, LLC

20

Best air delivery practices

Arrange racks in hot aisle/cold aisle configuration Try to match or exceed server airflow by aisle

Get thermal report data from IT if possible Plan for worst case

Get variable speed or two speed fans on servers if possible Provide variable airflow fans for AC unit supply

Also consider using air handlers rather than CRACs for improved performance (to be elaborated on later)

Use overhead supply where possible Provide aisle capping (preferably cold aisles, refer to LBNL presentation for more details) Plug floor leaks and provide blank off plates in racks Draw return from as high as possible Use CFD to inform design and operation

Taylor Engineering, LLC

21

Air-side economizer

Taylor Engineering, LLC

22

Air-Side Economizer issues

Hygroscopic dust
LBNL

is doing some research on this

Design humidity conditions


See

following slides

Taylor Engineering, LLC

23

Design conditions at the zone

2005, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Reprinted by permission from ASHRAE Design Considerations for Data and Communications Equipment Centers. This material may not be copied nor distributed in either paper or digital form without ASHRAEs permission.

Taylor Engineering, LLC

24

San Francisco
40
80

.020

75

.019

.018

PSYCHROMETRIC CHART
Normal Temperature
I-P Units

35

.017
75

70

WE TB

.016
UL BT EM PE RA

.015
TU R E-

.014
HUMIDITY RATIO - POUNDS MOISTURE PER POUND DRY AIR

16 FEET
BAROMETRIC PRESSURE: 29.904 in. HG

30

65

Upper Allowed Humidity Limit


14.0

70

.013

.012

San Francisco Climate Data Bins with Data Center Guideline Zones
Weather Hours 360 to 321 320 to 281 280 to 241 240 to 201 200 to 161 160 to 121 120 to 81 80 to 41 40 to 1
55

60
65

UME VOL

.011

25

.010

Design Target
60

.FT. - CU

.009

PER D LB.

20

50

% 90

% 80

55

X
Class1;Recommend

25

.008

A RY IR

.007

45

% 70

50

.006

Negligible time of possible concern for humidification


10

15

40
45

% 60

Class1;Allow
15%

.005

35
40
13.0

50%

40%

NEBS;Recommend

.004

35

.003
30%

30

20%

.002
E HUMIDITY

Lower Allowed Humidity Limit (20%RH)


.001 65 70 75 80 85 90 95 100

10% RELATIV

30

35

40

45

50

55

60

Chart by: HANDS DOWN SOFTWARE, www.handsdownsoftware.com

DRY BULB TEMPERATURE - F

Taylor Engineering, LLC

25

Los Angeles
40
.020

80
75 .019

.018

PSYCHROMETRIC CHART
Normal Temperature
I-P Units

35
70

.017

75

WE

TB

.016

UL BT

EM PE RA TU

.015

RE

.014

BAROMETRIC PRESSURE: 29.808 in. HG

Upper Allowed Humidity Limit


60

.013

.012

Los Angeles Climate Data Bins with Data Center Guideline Zones
Weather Hours 396 to 353 352 to 309 308 to 265 264 to 221 220 to 177 176 to 133 132 to 89 88 to 45 44 to 1
55

65

.011

25

.010

Design Target
P .FT. - CU
60
.009

20

50

% 90
% 80
55

X
Class1;Recommend

25

.008

.007

45

% 70
% 60
45
50%
40%
30%
20%

50

.006

Only a few hours of possible concern for humidification


10

15

40

Class1;Allow
15%

.005

35

40

NEBS;Recommend

.004

35
30

.003

.002

10% RELATIV

E HUMIDITY

Lower Allowed Humidity Limit (20%RH)


.001 65 70 75 80 85 90 95 100

30

35

40

45

50

55

60

Chart by: HANDS DOWN SOFTWARE, www.handsdownsoftware.com

DRY BULB TEMPERATURE - F

Taylor Engineering, LLC

HUMIDITY RATIO - POUNDS MOISTURE PER POUND DRY AIR

105 FEET

30

65

70

14.0 UME VOL ER D LB. A RY IR

13.0

26

Sacramento
40
80
75 .019 .020 .018

PSYCHROMETRIC CHART
Normal Temperature
I-P Units

35
70

.017

75

WE T

.016

BU LB TE MP ER AT

.015

UR

E-

.014

BAROMETRIC PRESSURE: 29.893 in. HG

Upper Allowed Humidity Limit


14.0

70

.013

.012

Sacramento Climate Data Bins with Data Center Guideline Zones


Weather Hours 270 to 241 240 to 211 210 to 181 180 to 151 150 to 121 120 to 91 90 to 61 60 to 31 30 to 1
55

60

65
25

.011

.010

Design Target
60
.009

20

50

% 90

% 80
45

55

X
Class1;Recommend

25

.008

.007

% 70

50

.006

Negligible time of possible concern for humidification


10

15

40

% 60
45
50%
40%

Class1;Allow
15%

.005

35

40

NEBS;Recommend

.004

35
30
30%
20%
10% RELATIV E HUMIDITY

.003

.002

Lower Allowed Humidity Limit (20%RH)


.001 60 65 70 75 80 85 90 95 100

30

35

40

45

50

55

Chart by: HANDS DOWN SOFTWARE, www.handsdownsoftware.com

DRY BULB TEMPERATURE - F

Taylor Engineering, LLC

HUMIDITY RATIO - POUNDS MOISTURE PER POUND DRY AIR

26 FEET

30

65

UME VOL .FT. - CU PER D LB. A RY IR

13.0

27

Lower humidity limit

Mitigate electrostatic discharge (ESD)

Recommended procedures

Personnel grounding Cable grounding Grounding wrist straps on racks Grounded plate for cables Grounded flooring Servers rated for ESD resistance Telecom industry has no lower limit The Electrostatic Discharge Association has removed humidity control as a primary ESD control measure in their ESD/ANSI S20.20 standard

Recommended equipment

Industry practices

Humidity controls are a point of failure and are hard to maintain Many data centers operate without humidification This needs more research Old technology not found in most data centers It is best to segregate these items rather than humidify the entire data center
28

And for some physical media (tape storage, printing and bursting)

Taylor Engineering, LLC

ESD control: floor grounding

Image from Panduit, reprinted with permission

Taylor Engineering, LLC

29

Water-Side Economizer
Integrated

Heat Exchanger in series with chillers on CHW side

Economizer Summary
Air-Side Economizers

Water-Side Economizers

Provides free cooling when dry-bulb temperatures are below 78F-80F. May increase particulates (LBNL research indicates this is of little concern). Should be integrated to be most effective. Improves plant redundancy! Can work in conjunction with water-side economizers on data centers! Need to incorporate relief.

Provides low energy cooling when wet-bulb temperatures are below 55F-60F. Avoids increased particulates (and low humidity if that concerns you). Should be integrated to be most effective (see previous slide). Improves plant redundancy! Can work in conjunction with air-side economizers on data centers!
31

Both are proven technologies on data centers!


Taylor Engineering, LLC

A case study of two designs

Collocation facility in the Bay Area Side by side designs in same facility over two phases Motivation for the second design was to reduce cost

Case study was developed by Lawrence Berkeley National Laboratory (LBNL)

Data Centers 8.1 and 8.2

Both sections at ~30% build-out during monitoring

Taylor Engineering, LLC

32

A tale of two designs: overview

Phase 1 Data Center (8.1)


Phase 2 Data Center (8.2)


26,200 ft2 27 W/ft2 design Traditional under-floor design with CRAC units Air-cooled DX Humidity controls (45%-55%)

73,000 ft2 50 W/ft2 design Under-floor supply from central AHUs with CHW coils Water-cooled plant Air-side economizers No humidity controls

Taylor Engineering, LLC

33

A tale of two designs: a closer look


Normalized efficiency metric: cooling
1.60 1.40 1.20
Normalized energy

kWcooling _ systems kWservers


~1/4 of the normalized energy

Phase 1 Data Center (8.1) Phase 2 Data Center (8.2)

1.00 0.80 0.60 0.40 0.20 Computer Loads UPS Losses HVAC Lighting

Data normalized to computer loads


Taylor Engineering, LLC 34

A tale of two designs: results

Phase 1 Data Center (8.1) Around 2x the HVAC installed cost ($/ft2) Around 4x the energy bills (when normalized to server load) Acoustical problems Higher maintenance costs Lost floor space in data center due to CRACs

Phase 2 Data Center (8.2) Preferred by the facility operators and data center personnel

Taylor Engineering, LLC

35

Two data centers: summary

What made the difference?


Airside

economizers No humidity controls Water-cooled chilled water system AHUs instead of CRAC units

Taylor Engineering, LLC

36

Custom CRAH Unit (Large)

Taylor Engineering, LLC

37

Example CRAH Unit Comparison


Model Budget Cost Number of units net total cooling (btuh) net sensible (btuh) sensible (tons) CFM SAT airside dT Internal SP no. fans fan type no. motors HP/motor total HP BHP/motor Unit BHP unit width depth height filter type Water PD (ft) CHW dT GPM Total GPM Total BHP Option 1 Option 2 Std CRAC Custom Model 1 Custom Model 2 $ 16,235 $ 23,000 $ 41,000 21 13 4 434,900 410,000 841,000 397,400 399,000 818,000 33.1 33.3 68.2 16,500 25,000 50,000 49.90 59.30 59.00 25.10 15.70 16.00 2 0.8 0.8 1.8 1.8 3 3 2 Centrifugal Plenum Plenum 1 3 2 15 5 15 15 15 30 15 4.7 11.5 15 14.1 23 122 122 122 35 36 72 76 156 168 ASHRAE 20% MERV 13 MERV 13 13.5 ft 11.1 11.1 14F 20 20 66.80 44.00 88.00 1,403 924 66% 315 275 87%

Taylor Engineering, LLC

38

Example CRAH Unit Comparison

34% less water flow 13% less fan energy

More if you consider the supply air temperature and airflow issues

Excess fan capacity on new units 36% higher cost for units, but

Fewer piping connections Fewer electrical connections Fewer control panels No need for control gateway Can use the existing distribution piping and pumps (case study) Can use high quality sensors and place them where they make sense

Possibly less turbulence at discharge?

Taylor Engineering, LLC

39

Air cooling issues

Limitations on the data densities served (~200w/sf)


Air delivery limitations Real estate Hot aisles are approaching OSHA limits

Working conditions

Costly infrastructure High energy costs Management over time Reliability


Loss of power recovery Particulates

Taylor Engineering, LLC

40

Take Aways

Use air- or water-side economizers where possible Consider personal grounding in lieu of humidification Consider AHUs as an alternative to CRACs Consider VSDs on fans, pumps, chillers and towers Refer to ASHRAE, LBNL and Uptime Institute for more recommendations

Taylor Engineering, LLC

41

State of the future: cooling with liquid

Taylor Engineering, LLC

42