february 26, 2009 · ch 2, 200 9 chandrakant d. patel, hp laboratories 19 100 w, 200-300 w/cm2...

IEEE Santa Clara Valley Chapter, CPMT SocietyFebruary 26, 2009

www.cpmt.org/scv

© 2008 Hewlett-Packard Development Company, L.P.The information contained herein is subject to change without notice

Sustainable IT EcosystemComponents and Packaging Implications

Chandrakant PatelHP Fellow and DirectorSustainable IT Ecosystem Laboratory

Objective and OrganizationIEEE CPMT Talk, February 26, 2009

Components and Packaging Technology necessitates systemic viewpoint

Sustainable IT Ecosystem

• IT ecosystem can enable sustainable transformation• Deconstructing conventional business models through IT services

• Need based provisioning of resources

− Data centers at the core, billions of service oriented client devices at the edge• Universal accessibility will require reducing the Total Cost of Ownership

• TCO reduction will require “cradle to cradle” least energy, least material solutions

Components and Packaging design must take a system perspective

• End to End Design− Least Energy

• Examining the performance of the ensemble

− Least Material

• Through lifecycle engineering and management

2 2 March 2009, Chandrakant Patel


www.cpmt.org/scv

3 2 March 2009

Sustainable IT Ecosystembillions of service oriented client devices and thousands of data centers….to deliver net positive impact

Sustainable IT EcosystemApproach

• Next Generation Infrastructure - City 2.0

− Supply and Demand Side Management to meet the needs of the inhabitants

• Supply Side:

− Need to design the physical infrastructure with lifecycle engineering – energy required in extraction, manufacturing, operation and reclamation – in mind, so that the embedding of available energy into the built environment can be minimized

− Need to utilize local resources to minimize destruction of available energy in transmission, construction of transmission infrastructure, etc

• Demand Side:− Provision fundamental resources based on the needs of the user

− Can available energy from 2nd law be used to represent resources

• Use the IT Ecosystem to enable supply and demand side management based

4 2 March 2009


www.cpmt.org/scv

5 2 March 2009

Delivering Net Positive Impact through Supply and Demand Side Management • Deconstruct conventional business models and

replace with lower carbon IT services − Advantage of scale when billions utilize IT to

address their fundamental needs and improve quality of life

• Transformation necessitates− Reducing the cost of IT for universal accessibility

• Reducing TCO necessitates addressing sustainability with an end to end perspective

• Use the IT ecosystem to enable need based provisioning of resources across all ecosystems− Transformation necessitates

• pervasive sensing, knowledge discovery, and control

• Key Enablers:

− Tools & Unifying Metric

− Return to fundamentals of Physical Engineering in combination with Computer Science

− Human capital trained in the fundamentals –multidisciplinary curriculum

Life-Cycle Design

… Power Transport Water WastePolicy-Based Control & Operation

Knowledge Discovery, Data Mining, Visualization

Scalable & Configurable Resource Microgrids

Pervasive Sensing Infrastructure

City 2.0 Architecture


www.cpmt.org/scv

7 2 March 2009

extraction operationmanufacturing End of Life

Tools for Sustainable Transformationmetrics for least energy, least material ecosystems

8 2 March 2009 Chandrakant Patel, [email protected]

Approach2nd Law of Thermodynamics

• Can a measure of the total exergy or available energy destroyed across a product’s lifetime (“lifetime exergy”) be a measure of the environmental sustainability?

• Can we build a “hub” of exergy data to enable lifetime exergy analysis for a given product?

Extraction

exergy toproducematerials

exergy tofabricate and assemble components

exergy fortransportation

operationmanufacturing

exergy forinstallation

exergy to power electronics and thermal managment

exergy forwastemitigation

End of LIfe

exergy tode-install anddecommision

exergy todisassembleand/orrecondition for safe disposal or recycle

time

exergy forwastemitigation

Joules of Exergy consumed becomes the currency of the Sustainability Age


www.cpmt.org/scv

Sustainability HubCommunity-Owned Resource for Knowledge Sharing

• Influence, gather, and scale knowledge within the global sustainability community to enable an open-source service

Susta inability Hub and Service

• Community driven “ecopedia”

• Robust, consistent, verifiable data

Education and Outreach

Community-driven databases

Data and Context

Calculations and linkages

Information Representation

Standards

Reusable data and metrics

Scale the knowledge that we have created

Gather knowledge that we don’t have

Influence with the knowledge we create

Susta inability Hub and Service

• Community driven “ecopedia”

• Robust, consistent, verifiable data

Education and Outreach

Community-driven databases

Data and Context

Calculations and linkages

Information Representation

Standards

Reusable data and metrics

Scale the knowledge that we have created

Gather knowledge that we don’t have

Influence with the knowledge we create

310Process)by Consumed(Energy

usage)ater Indirect w and usageer Direct watin Consumed(Energy Waterindex

Exergy: Water example

Energy consumption Average per million gallons

Water Treatment 0.25MWh

Water distribution 1.3 MWh

Waste Water Treatment 2.5 MWh

Desalination 20 MWh

Total (excluding desalination)

~ 0.5 GWh to serve a city of a million per day

Average per capita usage in the USA: 100 gallons per day

Sharma et al., “Water Efficiency Management in Data Centers: Introducing a Water Usage Energy Metric”, International Conference on Water Scarcity, Global Changes and Groundwater Management Responses, Irvine, CA, December, 2008

(Ref. California Energy Comission)(ref. SANDIA/DOE)


www.cpmt.org/scv

11 2 March 2009

Power Micro-Grid Infrastructure

Datacenter

Ecosystem of Clients

Supply and Demand Side Management at City Scale

Sustainable IT Ecosystem Lab, HP Labs, 2008

Transport Micro-Grid InfrastructureWater Micro-Grid

InfrastructureEducation Infrastructure

• Least Lifecycle Exergy Destruction =>• Least Total Cost of Ownership

Medical&

Health Infrastructure

Role of the IT Ecosytem

12 2 March 2009

“The data center is the computer”

Power CoolingCompute

Flexible & Configurable Building Blocks

Sensing Infrastructure

Policy/SLA based Integrated Management

Data Analysis, Visualization, Knowledge Discovery

• End-to-end Management using requirements derived from service level agreements (SLAs) and a flexible infrastructure that can be closely monitored and finely controlled

Architecture for a Sustainable Data CenterEnd to end design and management

• Data Center Synthesis using end to end analysis Tools



www.cpmt.org/scv

Part 2Summary of Part 1

• Sustainable IT Ecosystem

• Enabling next generation of communities: City 2.0

• Role of the IT Ecosystem

• Services

• Need based Provisioning through supply and demand side management

• Introduced Unifying “Cradle to Cradle” Metric based on the 2nd Law of thermodynamics

• Supply and Demand side management centered around the core – a data center

• Physical representation of data center as a system: “data center is the computer”

Part 2: Components & Technology Implications

• Implications for components, and packaging technologies given the cradle to cradle systemic perspective of the ecosystem

− Reduce available energy consumed in operation

• Reduce available consumed by the devices during operation

• Reduce available energy consumed in supporting the operation of the devices e.g. cooling

− Reduce available energy embedded in devices• Introducing “exergo-thermovolume” metric for components

13 2 March 2009

14 2 March 2009

Ecosystem ViewImpact on Components and Technology

Wchip

Wsystem

Wpump

Wblower

Wpump Wcompressor

Wblower

Qchip

Qsystem

Qdata center + ∑W


k

ctcompm

pl

crbj

ri

devcp

dcG

WWWWWWW

QCOP

sup

Power Grid – Wensemble

Cooling Grid

Outside Air


www.cpmt.org/scv

15 2 March 2009

Coefficient of Performance of the Ensemble

k

ctcompm

pl

crbj

ri

devcp

dcG

WWWWWWW

QCOP

sup

Cooling Tower loop

Chiller Refrigerant loop

Chilled Water loop

Data Center CRAC units

Warm Water

Air Mixture In

QCond

Wcomp

Air Mixture In

Return Water

QEvap

Cooling tower wall are adiabatic, Q = 0

Makeup Water

Air Mixture Out

Wp

Wp

ondarypdchydronics WQCOP

sec

compW

chQ

chCOP

1

/)1(

2

3)1(

22nn

P

P

nmotor

nPrefm

compWp

ct

compdcctctct W

WQWQCOP

Patel, C.D., Sharma, R.K., Bash, C.E., Beitelmal, M, “Energy Flow in the Information Technology Stack: Introducing theCoefficient of Performance of the Ensemble”, ASME International Mechanical Engineering Congress & Exposition, November 5-10, 2006, Chicago, Illinois

Impact on Data Center Total Cost of OwnershipSustainability through end to end design and management => Least Cost

1$,12112

21,

$

depavgtotal

hardwareconsumedgridcriticaltotal ITSMRPULKLKftA

ftCost

Personnel, equipment, SW per rackBurdened power consumptionReal

Estate

J1 : capacity utilization factor, i.e. ratio of maximum design (rated) power consumption to the actual data center power consumption

K1 = F(J1): burdened power delivery factor, i.e. ratio of amortization and maintenance costs of the power delivery systems to the cost of grid power

K2 = F(J1): burdened cooling cost factor, i.e. ratio of amortization and maintenance costs of the cooling equipment to the cost of grid power

L1: cooling load factor, i.e. ratio of power consumed by cooling equipment to the power consumed by compute, storage and networking hardware (inverse of COPensemble)

Dep

reci

atio

n f

acto

rs

Patel and Shah, Cost Model for Planning, Development and Operation of a Data Centerhttp://www.hpl.hp.com/techreports/2005/HPL-2005-107R1.html



www.cpmt.org/scv

March 2, 2009

Chandrakant D. Patel, HP Laboratories 17

Components and PackagingComponents and Packagingstacked stacked packaging, integrated photonics, packaging, integrated photonics, high power densityhigh power density

1998 2006

10

100

1000Frequency

Device Size.

# of Devices

Voltage

Capacitance per device

per chip

2010

100

Heat Flux, W/cm2

PGAEnhanced Ball Grid Array

Flip Chip Array, Underfill

Flip Chip MCM, KGD Testing Issues

Flip Chip, Non Uniform Heat Distribution, Power Mgmt

Po

wer

, W

200

3d 3d pkgpkg

400

# of Cores

• High Power Density

• Temperature Control

2002

Optical Interconnect


Thermal Management Challengestacked devices: chip and package scale

Heat SinkHandheld chassis or forced air finned exchanger on a server

Q + Wrequired to remove heat

High power density microprocessor with stacked devices

Q

Q


www.cpmt.org/scv

March 2, 2009


100 W, 200-300 W/cm2

Active Cooling at Package LevelDemise of Passive Only Cooling Solution

Coolant

Wcp

Work required by the chip package

Interface

• Active given the power density

Interface

System Component & Package

March 2, 2009


Work Required at Package LevelHigh Power Density Chips

100 W, 200-300 W/cm2

Epoxy Glass Printed Circuit Board

Forced Air Heat Exchanger

Active Micromechanical Means

chip – heat sink interface Wchip-package

34 W of Power Required by Cooling Resources to remove 100 W

Fan Work: 4 W

~.008 m3/s, P of 75 Pa; wire to air: 15%

Thermo-electric Module

30 W for a 15 oC of temperature reduction between hot and cold side

TEC Interface Chip Package Work: 30 WWchip-package ~ 30 W (COP of 3 for T of15 C)


www.cpmt.org/scv

Flow Work & Thermodynamic Work

21 2 March 2009

Work required to remove heat:

• Flow & Thermodynamic Work

− Flow Work

− Volume flow and pressure drop

• volume flow, m3/s

• pressure drop, Pa

− Thermodynamic Work

• Temperature, °C Tinlet


Available Energy Consumed in Heat RemovalWork required to remove heat from chip and system enclosure

Heat Sink

Epoxy Glass Printed Circuit Board

Tin, Tout

Operation

idevcpsysoutsysinsyspa

devcp

devcp

sys

syssys WWTTCm

WW

QQ

W

QCOP sup,,,

sup

sup

)(

)(


www.cpmt.org/scv


Work Required at Rack Level

r

isys

isysr WWQCOP

Rack level blowers

Pblade Pa

Fig 5d. Volume Resistance

Thrblade, ºC/W

, m3/s

Fig 5c. Thermal Resistance

, m3/s

Fig 5e. Blower Characteristic Curve

Pstatic, Pa ζb,sys

ζb,sys

, m3/s

Blower Curve

Blade Volume Resistance Curve

Operating Point

System Flow Work determined from volume flow and pressure drop

System active cooling device e.g. blower

March 2, 2009


Computer Room Air Conditioning Unit (CRAC)

CRAC – AH 2: 92.2%

CRAC - AH 1: 94.2%

CRAC – AH 6: 27.7%

CRAC- AH 5: 36.7%

CRAC - AH 4: 38.8 %

CRAC- AH 3: 80.3%

crahbcrlcr WQCOP

Blower Characteristic Curve

Pstatic, Pa ζb,sys

ζb,sys

, m3/s

Example shown has a Chilled Water Air Handling Unit in the room – so we will refer to it as CRAC – AH (Contains Air Mover and a Chilled Water Coil)


www.cpmt.org/scv

March 2, 2009


k

ctcompm

pl

crbj

ri

devcp

dcG

WWWWWWW

QCOP

sup

COP of Ensemble

depavgtotalhardwareconsumedgrid

GG

criticaltotal ITSMRaPwrUCOP

KCOP

KmAm

Cost ,$,21

2

2

111,

$

• Representing L1 as inverse of COPG

Reducing the Cost of the Data CenterMaximizing COPG

26 2 March 2009

Wchip

Wsystem

Qsystem

Outside Air

References:1.Reducing Data Center Cost with an Air Economizer, IT@Intel Brief, August 20082.Best Practices in Energy Efficiency in Microsoft Data Centers, Feb 2008

System Consideration: Using outside air i.e. eliminating the chiller in the ensemble• Component design drives this upstream consideration• Need to revisit environmental specifications with lifecycle in mind• In face of challenges: high power density, optical interconnects, stacked die


www.cpmt.org/scv

Revisiting Environmental SpecificationsComponents need a “Damage Boundary” specification with respect to temperature

27 2 March 2009

Errors recoverable with retries

Amplitude, g’s

Change in Velocity, m/s

Short duration mechanical shock pulse used to establish vertical boundary

• pulse width established based on chassis fundamental frequency

Long duration mechanical shock pulse used to establish horizontal boundary

[3] Hedtke, L and Patel C.D., Damage Boundary Assessment of Hard Disc Drives, ASME WAM 1990

Drawing the analogy from late 1980s - “damage boundary” technique used to assess fragility of a disc drive to mechanical shock [3][4].

Hard Errors

1. Is there an analogous approach to establish temperature “damage boundary” for components?

2. What is the lifecycle impact?

[4] Newton, R.E., “Damage Boundary Revisited”, Technical Report 89-WA/EEP-24


Thermo-mechanical Models to assess Damageend of life prediction

Tin, °CTdrive,in

Tdrive core

Tdrive core,°C

Qdrive spindle


• Given inlet temperature, what is the core temperature of the spindle?

• Develop thermo-mechanical models to determine core temperatures of various electro-mech systems e.g.


www.cpmt.org/scv


Design ApproachLifecycle Based Engineering and Management

Extraction operationmanufacturing End of LIfe time

Cradle to Gate

Sensors for roadbeds: Performance requirement not changing with time, no need to upgrade for a long period • resistant to wide environmental requirements• ideally passive or low power requirement

Data Center, system chassis e.g. rack

Focus on embedded exergy

Gate to Grave

Focus on operational exergy

Servers for core data centers:Performance improvements allow more users per processors e.g. three year upgrade cycle for a microprocessor• Library of hybrid active-passive solutions

• scalable power and cooling solutions

Grave to Cradle

• appropriate material choices to enable reclamation• appropriate material choices to enable operation at elevated temperatures

Focus on embedded exergy

• Need for a systemic viewpoint in Components and Packaging− Computing is at the crossroads

• Early days of computing saw vertically integrated organizations building computing solutions

• Commoditization of compute hardware created an ecosystem of suppliers of hardware and software

− Computing solutions were integrated in place

− Data Center became the Computer

• Now there is an emergence of computing services – Cloud Computing

• Purveyed from data centers− Growth will come from billions who want to use services to improve the quality of life

− Growth will come from IT becoming seamless with physical infrastructures such as cities

− Component designs are affected by upstream and downstream considerations

30 2 March 2009

Key MessageIEEE CPMT Talk, February 26, 2009


www.cpmt.org/scv

31 2 March 2009

Enabled by a Sustainable IT Ecosystem

Joules: Currency of the Sustainability Ageecosystem: billions of handhelds and printers, thousands of data centers and print factories

City 2.0

february 26, 2009 · ch 2, 200 9 chandrakant d. patel, hp laboratories 19 100 w, 200-300 w/cm2...

Documents