Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
©2009 HP Confidential template rev. 12.10.09 1
Yuan Chen
Senior Research Scientist
Sustainable Ecosystems Research Group
Hewlett Packard Laboratories
Sustainable Data Center
2 © Copyright 2010 Hewlett-Packard Development Company, L.P.
HP LABS AROUND THE WORLD Global talent, local innovation
PALO ALTO
ST. PETERSBURG
HAIFA
BRISTOL
BEIJING
BANGALORE
SINGAPORE
3 © Copyright 2010 Hewlett-Packard Development Company, L.P.
HP LABS RESEARCH AREAS Innovation at every touch point of information
Sustainability
Information Analytics
Mobile & Immersive Experience
Printing & Content Delivery
Services
Networking
Intelligent Infrastructure
Cloud & Security
4
Sustainable Ecosystem Research Group
– Sustainable Data Center
• Integrated management of IT, power and cooling towards a Net-Zero energy data
center
– Resource Management as a Service
• Improve sustainability of urban infrastructure
• All the resources needed to run a building or a campus or a city – resources like light,
water, power, waste management -- are all orchestrated and optimized
5
Sustainability: A Holistic Perspective reducing overall available energy consumption
Available Energy consumed by IT
Available Energy saved by application of IT
Available Energy
6 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Number of Data Centers
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
Single Rack Computer
Room
Midsize DC Enterprise DC Large DC
Data Center Size
Nu
mb
er
of
insta
llati
on
s
Source: Gartner Group, March 2011
98%
The rest of the global economy
Total carbon emissions
2%IT industry
Data Centers at the Core Emerging data centers
7 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Supply Side:
− Lifecycle perspective
−available energy (exergy) required in extraction, manufacturing, operation and reclamation
−Utilize local sources of available energy
−Examine available energy in waste streams
Demand Side:
−Provision resources based on the needs
− Integrated management of IT, Power and Cooling
Supply and Demand Integration
Integrated Supply-Demand Management based on Service Level Agreements
Technical Approach
8 © Copyright 2010 Hewlett-Packard Development Company, L.P.
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible, Configurable and Efficient Micro-grids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics & Visualization
Data Center Scale Lifecycle Design
Sustainable Data Center Key Components and Key Elements
9
Sustainable Data Center Research Elements Lifecycle Design
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible and Efficient Resource Microgrids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics and Visualization
Data Center Scale Lifecycle Design
10 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Lifecycle Perspective Minimizing product lifecycle exergy consumption at design time
Extraction
exergy to produce materials
exergy to fabricate and assemble components
exergy for
transportation
operation manufacturing
exergy for installation
exergy to power electronics and thermal managment
exergy for waste
mitigation
End of LIfe
exergy to de-install and decommision
exergy to disassemble and/or recondition for safe disposal or recycle
time
exergy for Waste mitigation
“sustainable development is
development that meets the
needs of the present without
compromising the ability of
future generations to meet
their own needs”
the Brundtland Commission of the United Nations,
1987
11 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Kinetic Energy + Heat+ Waste Streams
Energy refers to quantity of energy, available energy quantifies the useful portion (or quality) of energy
Lifecycle Perspective Energy or Exergy as a measure
• The quantity of energy was conserved • The usefulness of energy was destroyed
12 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Server Lifetime Exergy Consumption
Extraction
5%
Manufacturing
13%
Transportation
1%
Operation (Power)
41%
Recycling
1%
Operation (Cooling,
incl. Infrastructure)
40%
Lifecycle Perspective Lifetime Exergy Advisor
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Handheld
Note
book
Desk
top
Bla
de
Serv
er
Data
Cente
r
Fra
ctio
n o
f Li
fecy
cle E
nerg
y
Operational
Embedded
Hannemann, C., et. al., “Lifetime Exergy Consumption as a Sustainability Metric for Enterprise Servers”, Proceedings of ASME Energy Sustainability, August 2008
13
Sustainable Data Center Research Elements Flexible Actuators
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible and Efficient Resource Microgrids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics and Visualization
Data Center Scale Lifecycle Design
14 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Data Center Infrastructure
15
Examples of Flexible Infrastructure Components
Power
• Power microgrids • Dynamic power control Cooling
• Cooling microgrids • Variable speed blowers • Adaptive vent titles
IT
• Virtualization • Dynamic resource allocation
Outside Air
Economizer
PV Power
Cooling Supply
Power Supply Data Center Demand
Water
Economizer
Grid PowerP
ow
er In
frastru
ctu
re
CRACs
IT
16
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible and Efficient Resource Microgrids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics and Visualization
Data Center Scale Lifecycle Design
Sustainable Data Center Research Elements Cross-Layer Sensing
17
Pervasive Cross-Layer Sensing
IT Hardware (Servers, etc.) Facilities Power and Cooling
Sea of Sensors Dozens of sensors monitor power and temperature within IT equipment
iPDU and Remote Sensing Hundreds of sensors monitor power and cooling resources at facility-level
IT Software (Applications and Services)
Service Health Reporter and Insight Control Software sensors monitor the health of applications and services
18
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible and Efficient Resource Microgrids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics and Visualization
Data Center Scale Lifecycle Design
Sustainable Data Center Research Elements Advanced Analytics
19 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Advanced Visualization
20 © Copyright 2010 Hewlett-Packard Development Company, L.P.
EXTRACTION OPERATION MANUFACTURING END OF LIFE
Annual Savings: 359 MWh (~10%); 179,580 G direct water; 287, 328 Kg CO2
Motif Mining applied to water and air cooled chiller ensemble
Data Mining Pattern mining of chiller ensemble
Ref: Deb Patnaik, Manish Marwah, Ratnesh Sharma, and Naren Ramakrishnan, Temporal
Data Mining Approaches for Sustainable Chiller Management in Data Centers, ACM
Transactions on Intelligent Systems and Technology, 2011,
Cluster Analysis
Multivariate Time Series Data
Event Encoding
Frequent Motif Mining
Symbolic representation
Transition-event sequence
Frequent motifs
Sustainability characterizationof frequent motifs
Other discrete data sources can be integrated
21
extraction operation manufacturing End of Life
IT
Power
Cooling
Flexible and Efficient Resource Microgrids
Pervasive Cross-layer Sensing
Integrated Control
Advanced Analytics and Visualization
Data Center Scale Lifecycle Design
Sustainable Data Center Research Elements Integrated Control Systems
Autonomous Control Dynamic Provisioning of IT and Cooling Resources based on the needs Three key goals
– Avoid doing nothing
– Do nothing well
– Meet SLAs
23 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Data Center Cooling Delivery and Distribution Cold and Hot
Ref: Chandrakant D. Patel, Cullen E. Bash, Ratnesh Sharma, Smart Cooling of Data Centers, in
ASME 2003 International Electronic Packaging Technical Conference and Exhibition
(InterPACK2003) , July 6–11, 2003 , Maui, Hawaii, USA
24 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Data Center Cooling Delivery and Distribution
Conventional control • a carry over from room level comfort cooling, • regulates the return air temperature to a given setpoint • Fans are typically fixed speed running continuously at 100% • Individual Air Handlers are self controlled and not coordinated • results in very cold air being supplied to the room ~ 12°C / 55°F on average due to mixing in the room
Room Level Temperature Sense Points
Room Chilled
Water Supply
AC Unit
Air Movers
Chilled Valve
T T
CW supply and control
TBlower &
Motor
Conventional Room Control
25 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Data Center Cooling Delivery and Distribution Conventional Room Control
26 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Data Center Cooling Delivery and Distribution Dynamic Zonal Cooling Control
Dynamic Zonal Cooling • Regulate cooling on the supply side of the air handlers. • Supply air cooling is regulated by sensing temperature at the intake of equipment racks • Fans use variable speed control • A much higher supply air temperature ~ 68-72°F, a much higher return air temperature ~ 85-95°F
Room Chilled
Water Supply
AC Unit
Air Movers
Chilled Valve
CW supply and control
TT
VFDVFD
Rack Level Temperature Sense Points
27 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Cooling Control in Data Center Level Zonal Effects of the CRAC units
CRAC 1 CRAC 3CRAC 2
CRA
C 6
CRA
C 5
CRA
C 4
Section A
Section B
Section C
Regions of Influence
Data Center Room
CRAC 1 CRAC 3CRAC 2
CRA
C 6
CRA
C 5
CRA
C 4
Section A
Section B
Section C
Regions of Influence
Data Center Room
TCI: Thermal Correlation Index: i
j
jiCRAC
TempTempCRACTCI
),(
Ref: C.E. Bash, C.D. Patel, R.K. Sharma, “Dynamic Thermal Management
of Air Cooled Data Centers,” the Tenth Intersociety Conference on
Thermal and Thermomechanical Phenomena in Electronic Systems
(ITherm ’06), 2006
28 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Cooling Control in Data Center Level
Thermal Correlation Index (TCI)
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
A1
.T1
A3
.T1
A5
.T1
A7
.T1
Aext
2.T
1
Aext
4.T
1
Aext
6.T
1
Aext
8.T
1
Aext
10
.T1
B2.T
1
B4.T
1
B6.T
1
Bext
1.T
1
Bext
3.T
1
Bext
5.T
1
Bext
7.T
1
Bext
9.T
1
C2
.T1
C4
.T1
C6
.T1
Cext
2.T
1
Cext
4.T
1
Cext
6.T
1
Cext
8.T
1
Cext
10
.T1
D2
.T1
D4
.T1
D6
.T1
E2
.T1
E4
.T1
E6
.T1
F1
.T1
F3
.T1
F5
.T1
F7
.T1
G1
.T1
G3
.T1
G5
.T1
G7
.T1
G9
.T1
crac1
crac2
crac3
crac4
crac5
crac6
crac7
crac8
29 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Cooling Control in Data Center Level Coordinated CRAC Control
R
A
C
K
s
Inlet Temps
PID SAT_SP
VFD_SP
Tmaster CRAC
PID SAT_SP
VFD_SP
Tmaster CRAC
PID SAT_SP
VFD_SP
Tmaster CRAC
Identification of Master Sensors
TCIs
Ref: C.E. Bash, C.D. Patel, R.K. Sharma, “Dynamic Thermal Management of Air Cooled Data Centers,” the Tenth
Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm ’06),
2006.
30 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Cooling Control in Data Center Level Coordinated CRAC Control Results
Conventional Mode
Dynamic Zonal Cooling Mode
HP Labs Data Centre
Palo Alto, CA
35% Energy savings Improved reliability Ref: C.E. Bash, C.D. Patel, R.K. Sharma, “Dynamic Thermal
Management of Air Cooled Data Centers,” the Tenth
Intersociety Conference on Thermal and Thermomechanical
Phenomena in Electronic Systems (ITherm ’06), 2006.
31 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Local Cooling Control Adaptive Vent Tile
AC Power
Microcontroller
DATA ACCESS LEVEL
1-Wire
AVT Base station
Actuator
Vent Tiles
12-24 VDC Power
Network
AVT Manager
Actuator Damper
32 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Local Cooling Control Vent Tile
0 20 40 60 80 100 12019
20
21
22
23
Time(min)
Deg
ree(
C)
0 20 40 60 80 100 1200
50
100
Time(min)
Ope
ning
(%)
TF1
TF2
TF3
VF1
VF2
VF3
•Inter-correlation •Nonlinearity
Ref: Zhikui Wang, Alan McReynolds, Carlos Felix, Cullen Bash, Christopher Hoover, Kratos: Automated Management of
Cooling Capacity in Data Centers with Adaptive Vent Tiles, in Proceedings of the 2009 ASME International
Mechanical Engineering Congress & Exposition (IMECE 2009), Nov 13-19, 2009, Lake Buena Vista, Florida.
33 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Local Cooling Control Adaptive MPC AVT control driven by Inlet temperatures
Temp reference
(Tref1, … TrefN)
AVT
Actuators
(1, …, M)
IT
Equipment
Temp
Sensors
(1, …, N)
Cooling
air flows MISO
Controller
Online Model
Estimator
Vent tile openings
(V1, … VM)
Temperatures
(T1, … TN)
Adaptive
Controller
Target
System
Temperatures
(T1, … TN)
1,...,1,0,ˆ)(
1,..,1,0,)(
,...,2,1,)(
)()1()(
)1(
)()1()()(
1
0
2
1
0
2
1
2
2
1
huiVikV
huiVikV
hpiTikT
ikBVikATikT
ikV
ikVikVTikTVJMinref
hu
iR
hu
iR
hp
iQref
V
Ref: Zhikui Wang etc., KRATOS: automated management of cooling capacity in data centers
with adaptive vent tiles, IMECE2009, November 13-19, Lake Buena Vista, Florida, USA
34 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Before Control After Control0
50
100T
ile O
peni
ngs
(%)
Before Control After Control
20
25
30
Tem
pera
ture
s(C
)
F G D E 44% Less Cooling Air Flow Needed
Local cooling on-demand delivery through Adaptive Vent Tile
Ref: Zhikui Wang etc., KRATOS: automated management of cooling capacity in data centers with adaptive
vent tiles, IMECE2009, November 13-19, Lake Buena Vista, Florida, USA
Local Cooling Control
35
Dynamic Resource Allocation in Virtualized Server Environments – Determine IT resources needed to maintain application performance
– Dynamically adjust IT resources based on needs
© Copyright 2010 Hewlett-Packard Development Company, L.P.
Management
Domain
Application
Controller 1
Node Controller 1
Application
Controller Q...
Node Controller 2 Node Controller S
Utilization Targets
Physical Server 1
...VM
App 1
VM
App Q
Physical Server 2
...VM
App 1
VM
App 2
Physical Server S
...VM
App 2
VM
App Q
Resource Allocations
Application
Controller 2
...
...
Performance
Target1
Performance
Target2
Performance
Target S
Ref: Zhikui Wang, Yuan Chen, Daniel Gmach, Sharad Singhal, Brian Watson, Wilson Rivera, Xiaoyuan Zhu, and Chris Hyser. AppRAISE: Application-level
Performance Management in Virtualized
Server Environment. IEEE Transaction on Network and Service Management, Vol. 6, No. 4, pp. 240-254, December 2009.
36
Dynamic Resource Allocation in Virtualized Server Environments Application controller
Application Controller
Feed-Forward
Controller
Feedback
Controller
rref
∆r
Client
r(Mean Response Time)
Node Controller
of VM V1
Node Controller
of VM V2
Node Controller
of VM VM
+
Workload
-
+ ...
Offline Perf.
Models
U
mdlU
refU
refMu
Client
Sensor
refu1
refu2
Ref: Zhikui Wang, Yuan Chen, Daniel Gmach, Sharad Singhal, Brian Watson, Wilson Rivera, Xiaoyuan Zhu, and Chris Hyser. AppRAISE: Application-level Performance Management in Virtualized Server Environment. IEEE Transaction on Network and Service Management, Vol. 6, No. 4, pp. 240-254, December 2009.
37
RUBiS Response Time
Dynamic CPU shares allocated to each component of the application
Dynamic Resource Allocation in Virtualized Server Environments
38
additional resource
number of resources
forecasted base workload
actual workload
workload not exceeding base workload
excess workload
Base Workload Predictor
CoordinatorPredictive Controller
Reactive Controller
Sever Pool 1 Server Pool 2
historical workload demand
Dynamic Capacity Provisioning of Server Farm
Combine predictive and reactive control to allocate resources at multi-time scales
• Identify long-term sustained patterns
• Proactively provision for base workload (every a few hours)
• Reactively handle any excess workload (every a few minutes)
Ref: Minimizing Data Center SLA
Violations and Power Consumption via
Hybrid Resource Provisioning. Anshul
Gandhi, Yuan Chen, Daniel Gmach, Martin
Arlitt, and Manish Marwah. Proceedings of
the Second International Green Computing
Conference (IGCC 2011), July 2011
39
Dynamic Capacity Provisioning of Server Farm
© Copyright 2010 Hewlett-Packard Development Company, L.P.
40
Integrated Management of IT and Cooling
– A data center is a dynamic thermal environment
• There are significant non-uniformities
• Some locations in a data center are much more efficiently cooled than others
– A significant opportunity for energy savings lies in the integrated
control and placement of IT workloads placement according to
cooling efficiency
© Copyright 2010 Hewlett-Packard Development Company, L.P.
41
Integrated Management of IT and Cooling LWPI: Local Workload Placement Index
CRAC
ventinletMINCRACCRACcracnodeinletrefnode TTSATSATTCITTLWPI )()( ,,
Thermal
Management
Margin
Air Conditioning
Margin
Hot Air
Recirculation
Ref: C. E. Bash and G. Forman, “Data center workload
placement for energy efficiency,” in Proc. of ASME
InterPACK Conf., Vancouver, Canada, 2007.
42
Integrated Management of IT and Cooling High-level architecture
Ref: Yuan Chen, Daniel Gmach, Chris Hyser, Zhikui Wang, Cullen Bash, Christopher Hoover, Sharad Singhal,
Integrated Management of Application Performance, Power and Cooling in Data Centers, in Proceedings of the
12th IEEE/IFIP Network Operations and Management Symposium (NOMS 2010), Apr 19-23, 2010, Osaka, Japan.
Facility Management
CRACSCRACS
IT Management
Pod
Controller
Application
Controllers
VMs Resource
Assignment
Nodes Resource
Consumption
App. Resource
Requests
Application
Measurements
Performance
ModelsDaffy
Floor Plan
Information
DSC
CRACS
Sensors Actuators
Blower Speed
Reference Air
Temperatures
Supply Air
Temperature
Thermal
Metrics
IT Power
State
VMs Resource
Consumption
Temperature
SensorsTemperature
SensorsTemperature
SensorsPhysical Node
VM
App.VM
App.VM
App. ...
Node
Controllers
VMs Resource
Allocation
Power
Control
• Dynamically adjust IT
resources based on needs
• Dynamically re-arrange
workloads
• Opportunistically turn on
and turn off idle servers
• Use LWPI to guide
workload placement
• Adjust cooling resources
in response to IT needs
43
Integrated Management of IT and Cooling
HPL Palo Alto data center
• 20 physical servers
• 9 in Rack 10
• 11 in Rack 34
• 35 Virtual Machines
• 2 interactive 3-tier apps
• 29 computational workloads
• 10 hour experiment
experimental Testbed
44
Integrated Management of IT and Cooling
Number of running servers Workload demand
Server power CRAC blower power
results
45
CDF of response times
Integrated Management of IT and Cooling results
Anticipated Savings Per Year (2MW data center)
Energy Costs CO2 Water
4.9GWh $492,000 2915 tons 4,700,000
Gallons
• 35% IT power reduction • 16% cooling power reduction • No performance degradation
Ref: Yuan Chen, Daniel Gmach, Chris Hyser, Zhikui Wang, Cullen
Bash, Christopher Hoover, Sharad Singhal, Integrated
Management of Application Performance, Power and Cooling
in Data Centers, in Proceedings of the 12th IEEE/IFIP Network
Operations and Management Symposium (NOMS 2010), Apr 19-
23, 2010, Osaka, Japan.
Data center
Outside air
Integrated Management of Supply and Demand HPL Palo Alto Data Center
PV micro grid
Cooling
infrastructure
power demand
Data center supply side Data center energy
demand side
Goals:
• Avoid doing nothing
• “Do nothing” well
• Keep the energy budget balanced
Chill
er
CO
P
Outside Air Temperature (°C) Load (kW)
47
Integrated Management of Supply and Demand
Execution
Supply-side Prediction
•Renewable power prediction •Cooling capacity prediction
IT Demand Prediction
IT Workload Planning • Integrate Supply and Demand Side • IT Demand Shaping • Power capping
Measurement
Verification
DC Operation Constraints: • Net-zero energy operation
• Net-zero costs operation
• Maximize use of renewable energy
• Minimize dependability on Grid
Dynamic IT Provisioning
Dynamic Cooling Provisioning
Prediction Planning
Verification and Reporting
6 12 18 240
0.2
0.4
0.6
0.8
1
1.2
Po
we
r(kW
)
Time(hour)
PV Supply
6 12 18 240
0.5
1
po
we
r (K
W)
time (hour)
critical workload
non-critical workload
cooling power
0 5 10 15 200.1
0.15
0.2
0.25
0.3
0.35
Time (hour)
Po
we
r (k
W)
Outside Air Cooling Available Capacity
6 12 18 240
0.2
0.4
0.6
0.8
1
1.2
po
we
r (K
W)
time (hour)
critical workload
non-critical workload
cooling power
renewable supply
1 2
3 4
48
0
200
400
600
800
1000
Supply-aware Supply-oblivious
En
erg
y (
KW
h)
Renewable Grid
Integrated Management of Supply and Demand
Pla
nnin
g
Imp
act
6 12 18 240
50
100
150
Time (hour)
Pow
er
(KW
)
critical WL
non-critical WL
cooling
renewable
6 12 18 240
50
100
150
Time (hour)
Pow
er
(KW
)
Supply-oblivious Supply-aware
Energy Consumption Cost
0%
20%
40%
60%
80%
100%
Recurring Power Energy Storage
Norm
ailz
ed
Cost
Supply-aware Supply-obivilous
-87% -86%
Work
load S
chedulin
g
Renewable and Cooling Aware Workload Management for Sustainable Data Centers. Zhenhua Liu, Yuan Chen, Cullen Bash, Adam Wierman, Daniel Gmach, Zhikui Wang, Manish Marwah, and Chris Hyser. Proceedings of the 12th International Conference on Measurements and Modeling of Computer Systems (SIGMETRICS 2012), June 2012. (to appear)
Towards The Design And Operation Of Net-Zero Energy Data Centers. Martin Arlitt, Cullen Bash, Sergey Blagodurov, Yuan Chen, Tom Christian, Daniel Gmach, Chris Hyser, Niru Kumari, Zhenhua Liu, Manish Marwah, Alan McReynolds, Chandrakant Patel, Amip Shah, Zhikui Wang, and Rongliang Zhou. Proceedings of the 13th Intersociety Conference on Thermal and Thermomechanical Phenomena in Electronic Systems (ITherm 2012), May 2012.
49 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Digester and Biogas Handling
Power Generation
Cooling Infrastructure
Manure
Biogas
Hot Water
Steam
Power
Data Center
Manure Handling
Dairy Farm
Waste Heat
Effluent Handling and Storage
Sustainable IT Ecosystems Server Farm at the Dairy Farm
Ref: Ratnesh K. Sharma etc, " Design of Farm Waste-driven Supply Side Infrastructure for Data
Centers," Proceedings of the ASME 2010 4th International Conference on Energy Sustainability, May
17-22, 2010.
How could a hypothetical farm of 10,000 dairy cows fulfill the
power requirements of a 1MW data center ?
50 © Copyright 2010 Hewlett-Packard Development Company, L.P.
Sustainable Data Center
Ecosystem of Clients
Data Centers at the Core for management of resources at community scale
Water Transport
Power
Education Healthcare
51 © Copyright 2010 Hewlett-Packard Development Company, L.P.
City 2.0 Enabled by a Sustainable IT Ecosystem
Traditional Ecological Engineering
Life-Cycle Design
…
Power
Transport
Water
Waste
Policy-Based Control & Operation
Knowledge Discovery, Data Mining, Visualization
Scalable & Configurable Resource Microgrids
Pervasive Sensing Infrastructure, Aggregation, Dashboard
Design
Mgmt
52 © Copyright 2010, Hewlett-Packard Development Company,
L. P.
Research Opportunities and Challenges
End-to-end Optimization Analytics for Sustainability Multi-scale Adaptive Control Integrated Control of Power, Cooling and IT Integrated Supply and Demand Management Beyond IT
53
Acknowledgement
© Copyright 2010 Hewlett-Packard Development Company, L.P.
The work presented in this talk has been contributed by the team of Sustainable Ecosystems Research Group at HP Labs, students and professors at CMU and Caltech:
Chandrakant Patel, Cullen Bash, Martin Arlitt, Yuan Chen, Tom Christian, Kiara Corrigan, Daniel Gmach, Chris Hyser, Niru Kumari, Geoff Lyon, Manish Marwah, Alan McReynolds, Amip Shah, Zhikui Wang, Rongliang Zhou, Sharad Singhal, Anshul Gandhi, Zhenhua Liu, Adam Wierman