Using Computational Fluid Dynamics (CFD) for improving cooling system efficiency for Data centers Data Centre Best Practises Workshop 17 th March 2009.

Slides:



Advertisements
Similar presentations
CFD ANALYSIS OF CROSS FLOW AIR TO AIR TUBE TYPE HEAT EXCHANGER
Advertisements

Data Center Design Issues Bill Tschudi, LBNL
Clair Christofersen Mentor – Aaron Andersen August 2, 2012
AIR DISTRIBUTION (Additional information. Also see Chapter 18) General The proper delivery of air for heating, cooling, or ventilation is a crucial part.
1 Application of for Predicting Indoor Airflow and Thermal Comfort.
Heat Exchanger Network Retrofit
Hongjie Zhang Purge gas flow impact on tritium permeation Integrated simulation on tritium permeation in the solid breeder unit FNST, August 18-20, 2009.
Using Copper Water Loop Heat Pipes to Efficiently Cool CPUs and GPUs Stephen Fried President Passive Thermal Technology, Inc.
Improving and Trouble Shooting Cleanroom HVAC System Designs By George Ting-Kwo Lei, Ph.D. Fluid Dynamics Solutions, Inc. Clackamas, Oregon.
Computational Fluid Dynamics
The Three Tiered Philosophy
DATA ONE, EUROPEAN OPERATIONS CENTRE, ST. PETERSBURG, RUSSIA Presentation of Design Proposal.
STEM Center Delaware County Community College – Media, PA Thesis Final Presentation Dan Saxton Mechanical Option.
Cooling Product Positioning
OpenFOAM for Air Quality Ernst Meijer and Ivo Kalkman First Dutch OpenFOAM Seminar Delft, 4 november 2010.
1 ISAT Module III: Building Energy Efficiency Topic 6:Stead-State Building Loads z Fabric Loss z Ventilation Loss z Environmental Temperature z Steady-State.
1 Meeting ASHRAE Fundamentals, Standard 55 & 62.1 with Chilled Beams Displacement Ventilation.
Extended Surface Heat Transfer
Chapter 2: Overall Heat Transfer Coefficient
Project Motivation: Opportunity to explore building efficiency technology and the engineering design process Improving the thermal efficiency will save.
Jennifer Tansey 12/15/11. Introduction / Background A common type of condenser used in steam plants is a horizontal, two- pass condenser Steam enters.
Alan H Huber Physical Scientist; PhD, QEP NOAA, ASMD, in partnership with the US EPA, National Exposure Research Laboratory, RTP, NC, USA THE 5TH ANNUAL.
MuCool Absorber Review meeting FermiLab, Chicago 21 – 22 February 2003 Fluid Flow and Convective Heat Transfer Modelling by Wing Lau & Stephanie Yang Oxford.
Jordanian-German Winter Academy 2006 NATURAL CONVECTION Prepared by : FAHED ABU-DHAIM Ph.D student UNIVERSITY OF JORDAN MECHANICAL ENGINEERING DEPARTMENT.
University of South Carolina FCR Laboratory Dept. of Chemical Engineering By W. K. Lee, S. Shimpalee, J. Glandt and J. W. Van Zee Fuel Cell Research Laboratory.
Enclosure Fire Dynamics
Effect of Rack Server Population on Temperatures in Data Centers CEETHERM Data Center Laboratory G.W. Woodruff School of Mechanical Engineering Georgia.
Dynamic Reduced-order Model for the Air Temperature Field Inside a Data Center G.W. Woodruff School of Mechanical Engineering Georgia Institute of Technology.
 Site  Requirements  Local Resources  Initial layout ideas  Brief material selection  Supply options.
RF-Accelerating Structure: Cooling Circuit Modeling Riku Raatikainen
Best Practices in HVAC Design/Retrofit
School of Civil Engineering Integrating Heat Transfer Devices Into Wind Tower Systems to provide Thermal Comfort in Residential Buildings John Kaiser S.
Gyeongsang National University Hanshik Chung. CONTENTS  Background of Study  Introduction & Objective  Results and discussion  Conclusions.
Evaporation Slides prepared by Daene C. McKinney and Venkatesh Merwade
Air Conditioning and Computer Centre Power Efficiency The Reality Christophe Martel Tony Cass.
Address: Washington street 40 B-1050 Brussels Belgium Tel: Fax: rehva Federation of European.
Sensor-Based Fast Thermal Evaluation Model For Energy Efficient High-Performance Datacenters Q. Tang, T. Mukherjee, Sandeep K. S. Gupta Department of Computer.
Data centre air management Case studies Sophia Flucker.
A particle-gridless hybrid methods for incompressible flows
Thermal-aware Issues in Computers IMPACT Lab. Part A Overview of Thermal-related Technologies.
Dealing with Hotspots in Datacenters Caused by High-Density Computing Peter Hannaford Director of Business Development EMEA.
Thermal Aware Data Management in Cloud based Data Centers Ling Liu College of Computing Georgia Institute of Technology NSF SEEDM workshop, May 2-3, 2011.
Virtual Room Simulator (VRSIM). Outline Introduction to VRSIM - What is CFD - What is VRSIM - VRSIM Application - Problem Analyzed by VRSIM Case Solved.
PRESENTATION OF CFD ACTIVITIES IN CV GROUP Daniel Gasser.
INTRODUCTION TO CONVECTION
Power and Cooling at Texas Advanced Computing Center Tommy Minyard, Ph.D. Director of Advanced Computing Systems 42 nd HPC User Forum September 8, 2011.
SAHPA ® South African Heat Pipe Association Energy Postgraduate Conference EPC2013, Aug 2013 iThemba LABS Theoretical modeling and experimental verification.
1 Teaching Innovation - Entrepreneurial - Global The Centre for Technology enabled Teaching & Learning D M I E T R, Wardha DTEL DTEL (Department for Technology.
Energy Savings in CERN’s Main Data Centre
Turbulence Models Validation in a Ventilated Room by a Wall Jet Guangyu Cao Laboratory of Heating, Ventilating and Air-Conditioning,
1 A Self-Cooled Lithium Blanket Concept for HAPL I. N. Sviatoslavsky Fusion Technology Institute, University of Wisconsin, Madison, WI With contributions.
Date of download: 5/27/2016 Copyright © ASME. All rights reserved. From: Experimentally Validated Computational Fluid Dynamics Model for a Data Center.
7/15/2002PP.AFD.09 1 of 43 Yaskawa Electric America Variable Frequency Drives In HVAC Applications.
1 PCE 2.1: The Co-Relationship of Containment and CFDs Gordon Johnson Senior CFD Manager at Subzero Engineering CDCDP (Certified Data Center Design Professional)
Lecture Objectives: Accuracy of the Modeling Software.
Computational Fluid Dynamics
1 Copyright © 2016, The Green Grid The webcast will begin shortly Today’s live session will be recorded.
CFD Computational Fluid Dynamics. What is CFD? CFD is a form of digitally testing the airflow through the internals of a building. Computational fluid.
Unit 2: Chapter 2 Cooling.
The Data Center Challenge
Flow mal-distribution study in cryogenic counter-flow plate fin heat exchangers Geet Jain1, Sharad Chaudhary1, Prabhat Kumar Gupta2, P.K. Kush2 1Institue.
INTRODUCTION : Convection: Heat transfer between a solid surface and a moving fluid is governed by the Newton’s cooling law: q = hA(Ts-Tɷ), where Ts is.
Chamber Dynamic Response Modeling
Data Center Research Roadmap
Date of download: 11/2/2017 Copyright © ASME. All rights reserved.
Lecture Objectives Discuss: Project 1 Diffuser modeling
Thermal analysis Friction brakes are required to transform large amounts of kinetic energy into heat over very short time periods and in the process they.
Maintenance at ALBA facility: Strategy for the conventional services
FBE03: Building Construction & Science
Blasch Precision Ceramics
Presentation transcript:

Using Computational Fluid Dynamics (CFD) for improving cooling system efficiency for Data centers Data Centre Best Practises Workshop 17 th March 2009 Shishir Gupta

↓ You are Here ↓ Data Centre Case Study – Geometrical Details Introduction to CFD CFD while designing of HVAC system CFD during installation of Data Centre CFD for maintenance of Data Centre – Feedforward System

Computational (having to do with mathematics & computation) Fluid Dynamics (the dynamics of things that flow) CFD is built upon fundamental physics equations: equations of motion and conservation. CFD applications range from numerical weather prediction to vehicular aerodynamics design. CFD applications are linked with advances in computing software and hardware. CFD software is characterized by the physical models in the software. Fine-scale CFD applications closely match the true geometry of the physical objects and processes being modeled. Introduction to CFD

Mathematics Navier-Stokes Equations Fluid Mechanics Physics of Fluid Fluid Problem Computer Program Programming Language Simulation Results Computer Grids Geometry Numerical Methods Discretized Form Comparison& Analysis CFDCFD What is CFD?

Some DangerousSafeSecurity SomeAllRepeatable Measured PointsAllInformation Small/MiddleAnyScale LongShortTime ExpensiveCheapCost ExperimentSimulation(CFD) Why use CFD?

reactor vessel - prediction of flow separation and residence time effects. Streamlines for workstation ventilation HVAC Chemical Processing Hydraulics  Chemical Processing  HVAC(Heat Ventilation Air Condition)  Hydraulics  Aerospace  Automotive  Biomedical  Power Generation  Sports  Marine Where use CFD?

 Chemical Processing  HVAC  Hydraulics  Aerospace  Automotive  Biomedical  Power Generation  Sports  Marine Temperature and natural convection currents in the eye following laser heating. Aerospace Automotive Biomedicine Where use CFD?

Flow around cooling towers Marine SportsPower Generation  Chemical Processing  HVAC  Hydraulics  Aerospace  Automotive  Biomedical  Power Generation  Sports  Marine Where use CFD?

↓ You are Here ↓ Data Centre Case Study – Geometrical Details Introduction to CFD CFD while designing of HVAC system CFD while installation of Data Centre CFD for maintenance of Data Centre – Feedforward System

CFD Case Study for Data Centre

Introduction to the Case Study Case Study is taken from one of the project that we did for a Data Centre in India The case study includes what we did for the client also extends it for what could have been done for the same project using CFD There were two software applications used for the project : OpenSource CFD platform of OpenFoam and commercial CFD package of Fluent Both packages produced about the same results, in this presentation the results from OpenFoam are being shown

Case Description The analyzed Data Centre is rectangular with of area 516m 2 and height 3.35mt Cooling is to be provided using raised flooring layout and demarcation is done for Cold Aisle and Hot Aisle The sources of heat gain inside the data centre are listed below: – Heat gain through exterior walls accounting for thermal resistance of the wall – Heat gain from Server Racks, 154 Server racks each providing about 8 KW combine to about 1.26 MW Three fans of about 500CMH were assumed to transport air from cool aisle to hot aisle in each rack unit (Since detailed blade specification is not known)

HVAC System Specification 10 CRAC units, 1 Standby Specification: – Each CRAC unit of 30,585 CMH – Cooling capacity of Each Rack is 150 KW – Temperature of supply air is 9.4 o C – Return Air opening area (On top surface): 2.23 m 2 Supply Air Diffuser (Cold Aisle) Specifications: – Dimension of 600mm X 600mm – 70% open area – 1 supply diffuser per rack (Total 154) – Supply air velocity can be controlled using under floor fan Return Air Diffuser (Hot Aisle) Specification: – Dimension of 600mm X 600mm – 50% open area – Total no. of diffusers: 242

Objective of the Study To maintain recommended temperature by ASHRAE for Class 1data centre The recommended atmosphere is defined as: – Temperature of 20 o C - 25 o C – Relative humidity of 40% - 55 % – The allowed change in temperature should be less than 5 o C/hr

Recommended Operating Conditions

Design Parameters Number of CRAC’s Location of CRAC’s Velocity of supply air

↓ You are Here ↓ Data Centre Case Study – Geometrical Details Introduction to CFD CFD while designing of HVAC system CFD while installation of Data Centre CFD for maintenance of Data Centre – Feedforward System

Base Case Design Isometric View of the Designed Data Centre Server Racks False Flooring False Ceiling Supply Diffusers Return Diffusers CRAC Units (11 Nos.)

Case Study Cont… COLD AISLE Diffusers HOT AISLE Diffusers Server Racks CRAC Units (11 Nos.) Top View of the Designed Data Centre

CFD Simulation of Base Case Temperatures across Y-Z plane

Temperature Contour Temperature Profile at vertical planes along the racks and cold aisle.

CFD Simulation of Base Case Temperatures across X-Y plane

Temperature Contour Temperature Profile at Horizontal planes along the racks and cold aisle. Lets look at the mid-plane contour in more detail…..

Temperature Contour in Middle Plane The temperature contour at the Horizontal plane at the middle portion of the rack

CFD Simulation of Base Case Temperatures across X-Z plane

Temperature Contour Temperature Profile at the middle plane is showing most uneven distribution. Lets analyse the middle plane in detail

Temperature Contour in Middle Plane The temperature contour at the vertical plane at the middle portion of the rack

Velocity Vectors in Middle Plane The Velocity Vectors at the vertical plane at the middle portion of the rack

Conclusion from the base case CFD 1.The Average temperature on the rack surface at the cold Aisle side is 15 2.The temperature at Cold Aisle is varying from 12 to 17 3.The Average temperature on the rack surface at the Hot Aisle side is 27 4.The temperature at Hot Aisle is varying from 18 to 32 5.The simulation shows that a good number of servers are experiencing temperature well above and below the ASHRAE recommended temperature levels 6.Short circuiting of cold air is clearly visible in the simulation

Optimizing number of CRAC units & Supply Air Velocity 1.Maximum heat load : 154 X 8 = 1264 KW (1.26 MW) 2.Heat capacity of each CRAC : 150 KW 3.Minimum number of CRAC required: [8.4] = 9 4.The system was designed with 9 CRAC units and velocity of supply air was adjusted to avoid short circuiting and temperature stratification 5.In this case the velocity of 2.2 m/s is coming out to be higher 6.The simulation was conducted with velocity of 1.6, 1.7, 1.8, 1.9, 2.0 & 2.1 m/s 7.The results with 1.8 m/s showed best results

Temperature Distribution with 9 CRACs & 1.8 m/s The temperature contour at the vertical plane at the middle portion of the rack

Velocity Vectors with 9 CRACs & 1.8 m/s The Velocity Vector at the vertical plane at the middle portion of the rack

Results of improved design CFD 1.The Average temperature on the rack surface at the cold Aisle side is 16 2.The temperature at Cold Aisle is varying from 13 to 17 3.The Average temperature on the rack surface at the Hot Aisle side is 23 4.The temperature at Hot Aisle is varying from 19 to 29 5.Short circuiting of cold air is reduced to a substantial level, however still prevalent 6.The simulation shows that a most of the servers are experiencing temperature as recommended by ASHRAE

Conclusion Using Computational Fluid Dynamics the system was designed to reduce to 90% of original design, thus bringing about first cost saving of 10% in the capital cost. The new system uses less energy and produces better result than the initial design based on thumb rules

↓ You are Here ↓ Data Centre Case Study – Geometrical Details Introduction to CFD CFD while designing of HVAC system CFD during installation of Data Centre CFD for maintenance of Data Centre – Feedforward System

Case Description The capacity of this data centre of of 42 X 154 = 6,468 Server Blades 4,000 server blades are to be installed 1,000 servers are by Dell, 2,000 by IBM & 1000 by Sun The design variables are: – Number of CRAC units – Which CRAC unit should be operational – Location of Server Blades in the racking system – Velocity of supply air inlet

CFD Simulation Setup The power requirement of 3000 Server is minimum 713 KW – 5 CRAC (750KW) are minimum number of units which can provide the required tonnage The CFD simulation were conducted with various locations of Servers, CRAC’s and Supply air velocity The best result was found with following parameters: – Top Racks are empty – Alternative CRACs are operating – Velocity of Supply air is 1.2 m/s

CFD Simulation Results Server Positions in the Racks

CFD Simulation Results Operational CRAC’s

Temperature Distribution with 5 CRACs & 1.2 m/s The temperature contour at the vertical plane at the middle portion of the rack

Velocity Vectors in Middle Plane The Velocity Vectors at the vertical plane at the middle portion of the rack

Calibration during Installation Temperature Sensors The Result from CFD shall be compared with average reading shown by temperature and velocity sensors If there is any difference, the modeling shall be improved to arrive at the actual values.

↓ You are Here ↓ Data Centre Case Study – Geometrical Details Introduction to CFD CFD while designing of HVAC system CFD during installation of Data Centre CFD for maintenance of Data Centre – Feedforward System

Feedforward System Whenever capacity of the data centre is to be increased, the design parameters like number of CRACs and supply air velocity should be determined using CFD If the capacity ramp up is not that frequent than CFD simulation can be conducted at that stage to arrive at design parameters If ramp-up/ramp-down is very frequent then a custom made CFD code should be developed using OpenSource Libraries. This would enable data centre administrator to conduct CFD’s for his data centre and analyze various design options

Conclusion CFD can help design and operate the data centre HVAC system with optimum efficiency Thank You