Brief Discussion on the submitted posters to PRAGMA 18, and AD for PRAGMA 20 Computer Centre

11Computer Centre, The University of Hong KongComputer Centre, The University of Hong Kong

Brief Discussion on

the submitted posters to PRAGMA 18, and AD for

PRAGMA 20

Computer Centre

W.K. Kwan([email protected])

5th March 2010


AgendaSubmitted posters to PRAGMA 18Brief discussion on that of GeosciencesBrief discussion on that of TelesciencesHPC / Grid Computing facilities in HKUCall of collaborationPRAGMA 20 to be hosted by HKU in March 2011


8 posters submitted in PRAGMA 18

All from our HKU HPC users1 A knowledge-based weighting framework to boost the power of

genome-wide association studies and a tool to integrate large genotype datasets for studying complex diseases

Biochemistry

2 An ab initio study of metal ion solvation, abundance and atmosphericmobility before strong earthquakes.

Earth Science

3 Cosmological N-body simulations in modified gravity theories using GADGET-2

Physics

4 Investigation of Urban Land Use Change and Anthropogenic Heat Effect on Temperature Change by WRF Model in Guangzhou-Shenzhen, South China

Civil Engin.

5 Heat Loss from Street Canyons Mech. Engin.

6 Large-Eddy simulation of Pollutant Plume Dispersion in the UrbanCanopy Layer over 2D Idealized Street Canyons

Mech. Engin.

7 Large-Eddy Simulation of Natural Ventilation in Urban Environment Mech. Engin.

8 Large-Eddy Simulation of Pollutant Removal in Neutrally & Unstably Mech. Engin.


Geosciences (Civil Engin.)Dr. Ji ChenInvestigation of Urban Land Use Change and Anthropogenic Heat Effect on Temperature Change by WRF Model in Guangzhou-Shenzhen, South China

(previous work from the same group)Large-scale Water Resources ManagementEffects of land surface change on regional hydrologic cycleRiver flow modeling of the Mekong River Basin


Geosciences (Mech. Engin.)

Heat Loss from Street CanyonsLarge-Eddy simulation of Pollutant Plume Dispersion in the UrbanCanopy Layer over 2D Idealized Street CanyonsLarge-Eddy Simulation of Natural Ventilation in Urban EnvironmentLarge-Eddy Simulation of Pollutant Removal in Neutrally & Unstably stratified Street Canyons


Computational Solution to Air Pollution Physics & Chemistry in Dense & Compact Cities

W.C. Cheng, James O.P. Cheung, Tracy N.H. Chung, Chun-Ho Liu & Colman C.C. Wong

Department of Mechanical Engineering, The University of Hong Kong

Lilian Y.L. Chan, Frankie F.T. Cheung & W.K. KwanComputer Centre, The University of Hong Kong


Core Research Interests

Air pollution physics & chemistryGeophysical turbulenceScientific computing

with applications for dense & compact cities


Major Materials Reported to PRAGMA18

Air pollution physics & chemistry in urban areas are studied by computation fluid dynamics (CFD)

Large-eddy simulation (LES)Open-source CFD code OpenFOAM-15 or -16OpenMPI or MPICH for inter-processor comm. HPCPOWER2 Linux cluster of the Computer Centre, the University of Hong Kong

24 blades 64-bit dual quadcore 3 GHz Xeon2.3 Tera-FLOPS (Theoretical Rpeak)

Extending to GRIDPOINT128 blades 64-bit dual quadcore 2.53 GHz Nehalem10.3 Tera-FLOPS (Theoretical Rpeak)


Background & ObjectiveBuoyancy-driven flows are commonly found in the daytime atmospheric boundary layer (ABL)CFD LES was used to examine their contributions to the ventilation & pollutant removal in urban street canyons

Major FindingsHigher wind speed & turbulent kinetic energy (TKE) are found in unstable stratificationPollutant plume is a shallow layer over the street canyons in neutral stratification but is carried rapidly upward into the ABL in unstable stratificationSolar irradiation in the daytime ABL enhances pollutant removal from urban areas

Large-Eddy Simulation of Ventilation & Pollutant Removal in Neutrally & Unstably Stratified Street Canyons


Pollutant removal in neutral stratification

Pollutant removal in unstable stratification

More rapid pollutant plume rise in unstable stratification


Background & ObjectiveNatural ventilation is influenced by building envelope in our modern environmentsNatural ventilation behavior in low-rise structures was examined using CFD LES

Major FindingsRecirculating flow in front of the windward openingTurbulence intensities can be promoted by 50% at the openingsVentilation mechanism changes noticeably once neighboring buildings are includedA vast fresh air entrainment is observed on the windward roof levelReverse flow is occasionally observed at small building separation

Natural Ventilation in the Urban Environment


Fresh air entrainment

Occasional Reverse flow


Background & ObjectiveHeat loss from buildings to the surrounding is undesirable in wintertime due to the excessive energy consumptionCFD LES was used to compare the local heat transfer coefficient in the street canyons of different aspect ratios

Major FindingsOnly 1 primary recirculation is found in the wake interference regime (h/b = 1/4)Persistent flow reattachment & separation are found in the isolated roughness regime (h/b = 1/11)Maximum & minimum of local heat transfer coefficients coincide with the reattachment & separation points, respectively, on the street

Heat Loss from Street Canyons


Street Canyon flow in the skimming flow regimeh/b = 1/4 – Spatial contours of temperature

Prevalent wind

More heat loss on the windward side


Background & ObjectiveGaussian pollutant plume model has been widely used in the estimation of air pollutant distribution in the ABLCFD LES was used to examine the necessary modifications when the Gaussian plume model is applied in the urban canopy layer (UCL)

Major FindingsDispersion coefficient in the UCL follows the Gaussian-plume shape in generalRoof-level pollutant concentration is largely affected by the air entrainment / pollutant removal for the street canyons that differs from the Gaussian plume model markedly

Large‐Eddy Simulation of Pollutant Plume Dispersion in the Urban Canopy Layer over 2D Idealized Street Canyons


Fresh air entrainmentNon-Guassian plume shape


Higher Spatio-temporal ResolutionUse more refined/accurate CFD LESs to study the flow & pollutant transport especially the roles of small/transient turbulence scales

Multi-scale CouplingUse multi-block CFD LESs model to study the coupling among different length/time scalesHow large-scale atmospheric turbulence affects (purges) pollutants in street canyons at the bottom of the ABL

Chemistry & Micro-physicsStudy the transport of chemically reactive air pollutantsStudy the condensation/evaporation/phase-change of air pollutants

Comparison with Field MeasurementsUse latent heat/telescience/satellite images from the PRAGMA community to validate our CFD LES

Upcoming Work


Incorporation of satellite data (thanks to PRAGMA)


The University of Hong Kong


The University of Hong KongFounded in 1912Ranked 18th in "THES-QS World University Rankings 2007”Ranked 1st in “QS.com Asian University Rankings 2009”12,300 undergraduate students 9,900 postgraduate students10 faculties

Architecture, Arts, Business and Economics, Dentistry, Education, Engineering, Law, Medicine, Science, Social Sciences (a comprehensive University)


Central HPC Facility Advancement

Year Top500 System Rmax Rpeak

2009 --- System for HKU Grid Point (around 18000)2008 --- 192-Cores + 356-CPU 2376.3 4337.6 IBM Linux Cluster 2005 --- 356-CPU IBM Linux Cluster 1086.19 2033.62003 240t h 256-CPU IBM Linux Cluster 637.8 1433.62001 --- 64-CPU HP Linux Cluster --- 64.02000 --- 8-CPU Linux Cluster --- 12.81998 345th 48-CPU IBM SP2 22.2 30.71995 186th 32-CPU IBM SP2 6.6 8.51993 --- 8-CPU IBM SP1 --- 1.0


Central HPC Services at HKUSupport intensive computation for cutting-edge researches in the universityDeploy high performance and large scale computing facilities, and the latest software technologyUser training and consultation

Distribution of portfolio returns

0.00

0.05

0.10

0.15

0.20

0.25

-16 -12 -8 -4 0 4 8 12 16 20

$ Millions

Prob

abili

ty

Average return = $2 MVaR = $8 M5% of Occurrences


Our projects and researchers

http://www.hku.hk/cc/ccsystem/hkuhpcc/


HKU is not rich in

computing facilities, but strong in experience and

researchers


Grid computing activities of HKU


HKU has joined many High-speed Research NetworksHKU is at present a member of CNGrid, TWGrid, PRAGMA and APGrid PMA (due to HK’s special geographical and “political” position), and intends to further the development to establish a HKGrid through the HKU Grid Point development


Experience in multi-institutional / grid related activities: High-speed Research Network

HARNETInternet2China Science and Technology Network (CSTnet) Trans-Eurasia Information Network (TEIN2) Korea Research Environment Open Network (KREONET)CERNET (China Education and Research Network)


Grid Computing – Development & Collaborations

Local HPC clusters

Campus Grid

Regional Grid (HK Grid)

National Grid (CN Grid)

International Grid (EGEE, APGrid, PRAGMA)


China National GridTsinghua Univ.清华大学

Shandong Univ. 山东大学Computer Network Information Center,

Chinese Academy of Sciences(CAS)中国科学院计算机网络信息中心

Inst. of Appl ied Physi cs and Computational Mathemati cs

北京应用物理与计算数学研究所

Xian J i aoTong Univ.西安交通大学

The Univ. of Hong Kong香港大学

Univ. of Sci ence & Technology of China中国科学技术大学

Shanghai Supercomputer Center上海超级计算中心Huazhong Univ. of Sci ence & Technology

华中科技大学

ShenZhen Inst. of Advanced Technology,CAS中国科学院深圳先进技术研究院

IBM Cel l Grid PointIBM Cel l 网格结点


CNGRIDHKU is A Grid Point of CNGrid (China National Grid) since 2005 (the only one outside China mainland) Many HKU researchers have research collaboration with China partiesMany projects in CNGRID (863) are of national scale


中国国家网格 2008学术年会12月 18-20日召开上海超级计算中心·张江·上海

An afternoon agenda:13： 30—13： 50 宝钢公司：樊俊飞： CFD技术在宝钢的应用13： 50—14： 10 上海核工院：谢永城：高性能仿真平台在核电站设备分析法设计中的应用14： 10—14： 30 上海飞机设计研究所：张淼：大型客机气动设计问题14： 30—14： 50 上海 708研究所：王金宝： CFD技术在船舶设计中的应用14： 50—15： 10 上海交通大学空天学院：刘洪：飞机设计大平台介绍15： 30—15： 50 Simulia 公司：高尚武：ABAQUS最新发展技术15： 50—16： 10 上海交通大学船舶与建筑工程学院：尤云祥：高性能计算在船舶与海洋工程中的应用16： 10—16： 30 上海交通大学 CIMS研究所：范秀敏：基于网格的协同虚拟装配16： 30—16： 50 铁道部信息技术中心：邢智民：中国铁路网格货运信息综合应用系统研究与实现16： 50—17： 10 中科院计算所：邹永强：基于中国国家网格的高性能计算环境


PRAGMAPacific Rim Application and Grid Middleware Assembly (PRAGMA)A platform to report on the progress from focused working groups and also a venue to involve new collaborators in PRAGMA activities PRAMGA Workshop meets twice a yearHKU accepted as institutional member in PRAGMA 16 (March 23-25, 2009)At the moment, HKU is the only PRAGMA institutional member in HKSAR, and also the only institutional member in the southern part of Yangtze River in China.


PRAGMA


EGEEEnabling Grids for E-sciencEthe largest multi-disciplinary grid infrastructure in the worldmore than 140 institutions, about 300 sites in 50 countriesCurrently already EGEE-III, co-funded by the European Commission


EGEE (2)Centralized Grid research/operation consortium in EuropeTaiwan Academia Sinica Grid Computing (ASGC) invited in December 2005 to become the major Regional Operation Centre (ROC) in AsiaHKU CC Setup of EGEE (Enabling Grids for E-sciencE) Grid testbed at 2007, becomes site member of TWGrid VO


Certificate AuthorityAll grid sites and users require digital certificates for authenticationNow proposing to setup International Grid Trusted Federation (IGTF) Grid CA in HKU(interview next week)If OK, HKU issued certificates will be accepted by various grid consortiums


Looking ForwardWithout HPC, no incentiveWithout Grid, no prospective

Thank you !

Brief Discussion on the submitted posters to PRAGMA 18, and AD for PRAGMA 20 Computer Centre

Documents

Transcript of Brief Discussion on the submitted posters to PRAGMA 18, and AD for PRAGMA 20 Computer Centre