Sustainability Issues of Asia Pacific Region

Transcription

1 Sustainability Issues of Asia Pacific Region Simon C. Lin ( ) Institute of Physics, Academia Sinica, Taipei, Taiwan Simon.Lin@twgrid.org 27 Feb 2012, Taiwan CHAIN Workshop

2 Global Scientific Challenges Demand International Collaboration Internet and e-infrastructure have changed the way scientists do research No longer can the processing of the huge amount of scientific data and the production of scientific results be undertaken by a single scientist Unique moment in the history of Science: Scientists from different disciplines are sharing methodologies and tools Virtual Research Communities that span organizations around the world, through an integrated digital infrastructure that connects the trust and administrative domains of multiple resource providers, have become critical to the sciences of the 21st Centuries Enabling Opportunity to Do the Not Possible! 2

3 e-science in Asia Diversity Geographically large and culturally diverse in nature Level of scien6fic collabora6on o9en reflected by network connec6vity The region as a whole tradi6onally inexperienced in regional collabora6on There is no Policy Coordina6on mechanism in Research Data in Asia Why e- Science in Asia? Global infrastructure is establishing quickly Take advantage of sharing and collabora6on to bridge the gap between Asia and the world To address the challenge of regional coopera6on EGEE/EUAsiaGrid/EGI have helped building the unseen Regional CollaboraAon. One hopes many others will happen soon! 32

4 Asia Pacific Grid Initiative (APGI) in EGI-Inspire The partners from 9 Asian countries are acting as JRU hoping to sustain a permanent and scalable e-infrastructure The extension and diversity of AP region, which at the beginning was an issue, now is showing as a richness Consortium ready to participate to the new projects (CHAIN and EGI-Inspire) in a proactive way Able to engage collaborations not only between Grid experts, but also among scientists, at an international level Operation towards a sustainable, scalable, persistent e- Infrastructure, easy to use and embedded in the research environment is coordinated by APROC incorporated with EGI. Asia regional infrastructure is compliant with EGI structure and similar federated initiatives in the rest of the world (e.g. Latin America)

5 Potential Metrics to Measure the Success of Sustainability

6 European Grid Infrastructure (July 2011 and increase from Apr 2010) APGI are EGI partners Logical CPUs (cores) 248,424 EGI (+29.3%) 337,608 All PB disk and PB tape Resource Centres 329 EGI 346 All (+6.8%) 93 supporting MPI (+6.8%) Countries (+11.5%) 50 EGI 57 All 38 NGIs providing resources 26 National Operations Centres 12 NGIs in 4 Federated Operations Centres 1 EIRO providing resources (CERN) 19 countries in 4 non-european Operations Centres EGI-InSPIRE RI EGI-InSPIRE Update, Lyon

7 International Research Network Connections (IRNC) 7

8

9 e-science Networking in Asia Pacific Region Ø 6+ PB data in/out ASGC in 2010, 2011 Ø 11 Gbps reached PK-PAKGRID PK-NCP PH-ASTI- LIKNAYAN VN-IOIT-HN TH- NECTEC MY-UM- CRYSTAL ihep-cas CN-SDU- LCG2 HK-HKU IP Transit CERNET CSTNET HK APAN-JP NL JP 12.5G 2.5G 5G 622M SG TW 2.5G 10G US SINET WIDE I2 / GN2 JP-Tokyo- LCG2 JP-KEK1 JP-KEK2 JP-HIROSHIMA- WLCG2 IN-TIFR CN-BEIJING- LCG2 KREONET TWAREN/ TANET IN- VECC1 KR-KISTI- GCRT AARNET IN-VECC2 KR-KNU AU-ATLAS TW-FTT ASGC NYMU TW-NCUHEP TW-NTCU NTU TW-NIU ATLAS Sites CMS Sites ALICE Sites EUAsiaGrid Sites 5

10 Evolution of Grid in Asia CPU Utilization: ~ 2,200 times in 7 years 10 times in 7 months, 16 months, and 34 months time 3.1 M CPU-Hrs/month, 3M Jobs/Month 12.8 PB data In/Out of ASGC for WLCG in past 24 months Reliability is still an issue for most non-wlcg sites in Asia Registered User & Host: 774/1,789; 1,449/2,619 (APGridPMA) 7,260/10,688 Cores, 5.42/7.47PB Disk, 4PB Tape PH ID NZ TW IN KR PK JP SG AU KEK HK NCP KISTI TH PKU MY VN 6

11 TW: 40% JP: 26% + 8% AU: 12% KR: 4% IN: 4% PK: 2% MY: 2% 7

12 e-science Collaboration in Asia Pacific Discipline Applica+ons Partners Plan for DesktopGrid HEP ATLAS, CMS, ALICE TH, TW, CESNET, INFN X BioMedical Virtual Screening for Drug Discovery Avian Flu, Dengue Fever MY, TW, VN, CESNET, INFN X Pandemic disease analysis VN, FR BioinformaHcs Grid enabling phylogenehc inference SVM Parameter ophmizahon for predichon of Caspases Genome search to idenhfy T3SS effect SG, TW, VN, CESNET, INFN X Autodock ligand- receptor docking X Complex diseases studies Earth Science Disaster MiHgaHon on Earthquake ID, MY, PH, TH, VN, TW, CESNET, INFN X Comp Chemistry Chemical compound property analysis TH, TW, CESNET X Climate Change Weather simulahon, sea level rising ID, PH, TH, VN, TW Social Sci. Social SimulaHon TW, UK X EGI-InSPIRE RI

13 Application Repository Application Status: S1 (in consideration), S2 (running but not ported to glite yet), S3 (ported to glite, unavailable in EUAsia VO), S4 (available in EUAsia VO), S5 (ready for production)

14 CERN Mont Blanc Large Hadron Collider for High Energy Physics as Driver for Global e-infrastructure Geneva Basic physics research Asking and answering fundamental questions High Energy accelerator: Giant microscope (p=h/λ) Generate new particles (E=mc 2 ) Create conditions close to the Big Bang (1ns after) 4 les.robertson@cern.ch EGI-InSPIRE RI hfp://event.twgrid.org/chep2010/index.html 4

15 Grid Virtual Screening Service by AutoDock One-click job submission Submit the docking job to the Grid with just one click Visualize your job status View the best conformation of a simulation SG + DG Generate the histogram with a given energy threshold 16

16 Dengue Fever Data Challenge in 2009 Total number of completed docking jobs Estimated needed computing power Duration of the experiment Cumulative computing results Total Computing Recourses in EUAsia VO Number of used Computing Elements 300,000 4,167 CPU*days 60 days 42.5 GB 268 Cores 6 Collaborators: UPM, MIMOS, MY IAMI, VN; HAII, TH Cesnet, CZ; GRC, TW 17

17 GVSS2 17

18 e-science for Earthquake Disaster Mitigation Seismic Sensor Networks Collaborators: PH, VN, TW, ID, MY, TH Local Sensor & Observation Data Global/Regional Sensor Data High Resolution Source & Rupture Process Analysis Fast Reporting System Archive Earthquake Data Center (SeisGrid) Ref. Historical Events Data Archive Forward Simulation & Event Construction on Grid Risk Analysis & Reduction 23

19 Future Works Hazard Maps Achieving full process of quantitative seismic hazard assessment Collecting and analyzing event data Understanding fault characteristics in details Facilitating accurate simulation on seismic waves Assessing anticipated earthquake and potential damages by the correct seismic and engineering models Maps of disaster coverage, risk and also evacuation are pragmatic to better preparedness 7168 grid points for Taiwan SGT DB takes 45 days on 80x 8-core nodes, and 100TB output Near-surface velocity

20 The detailed Tsunami hazard map for the complete South China Sea region would require up to 300K scenarios! The maximum water- level rise.

21 Sources of Water Disaster Seasonal monsoon Typhoon Pacific Ocean Mei-Yu

22 Human Activities Changes Climate 35. Aerosols (local, regional) Inversion Regional Climate Changes Urban heat Island effects Ozone, Biomass burning (local, regional) Applications 1.Health impact 2. Scenario study moist cool dry warm Dust +air pollutant Biomass burning moist cool

23 $'$%6$(.$,17,*,($8,9.&$(.$,#*%*:&6'!!"#$%&'()*+&,%)&-#)!!"#$%&'$()*+,-.&$(.$ ".+%/,0120#')#3)'&2$,&4)1"+'#5+'&! 6&%2)3+7)"$'(,+()*+&,%)! /0$1%$)&.*+,-.&$(.$ " "#$%&'()*+,$)-*.,/$#00()*123,%4&')5! 6&%2)3+7)(+/&(+%)! 21'#3)*)&1(*+,-.&$(.$ " 805$4&20#')#3)/#514+9)1"+'#5+'&!!#(&*)! $-.&$(.$ 1%,4*)*5.$()%&.,-.&$(.$ " :'03*)2"+#,*;)+91+,05+'2;)&'()%05$4&20#') " :%0'-)(&2&)+914#,&20#')&'()(&2&)50'0'-!.&2&)/&12$,+()<*)0'%2,$5+'2%!.&2&)-+'+,&2+()<*)%05$4&20#'%!.&2&)-+'+,&2+()<*)%+'%#,)'+27#,=%! ) & ' a $ ' a $ ( % 2 # 4+ G* "! 3 2 c 2 a AB02")2"&'=%)2#)C05)D,&*E

24 Are We Anticipating a Paradigm Shift in Asia Pacific?

25 You ve Heard of Exponential Growth World Performance per Dollar Spent Doubling Time (months) Chip capacity (# transistors) Optical Fibre (bits per second) Data Storage (bits per sq. inch) Number of Years Gilder s Law (32X in 4 yrs) Storage Law (16X in 4yrs) Moore s Law (5X in 4yrs) Triumph of Light Scientific American. George Stix, January 2001 Induction: What are Grids and e-science? May 18th,

26 Actual Internet Map It looks amazingly similar to Neural Net in terms of Connectivity It s a Scale-Free Network Everything over IP Always-On Everywhere

27 Internet Success Story is Unique n It is based on a single TCP/IP protocol n The governance issue is about names and numbers q The reliable Domain Name Service based on root servers of GTLD and CCTLD q IP address allocation by National NIC n It is a distributed service managed by a small number of people n Sustainability relies on the service charge for names and numbers n Practically Infinite Bandwidth is possible via Dark Fibre and Lambdas, limited only by the service providers Simon C. Lin EGEE 08, Istanbul

28 Remember NGI and Its Vision? NGI, Next Generation Internet This Internet will be so pervasive, reliable and transparent that we'll all just take it for granted. It will be a seamless part of life--like electricity or plumbing. Hoping to create an Internet that s Fast, Always On, Everywhere, Natural, Intelligent, Easy, Trusted

29 Some Observations on Grid Computing n There are too many possible ways to integrate the compute and data resources in Grid e-infrastrucure n A single Grid middleware stack such as glite is ideal but it seems to be challenging now n However, critical mass still matters in terms of the market share of Grid middleware n Collaboration with common goal is the foundation of sustainability n It is still a leapfrog opportunity in scientific research in Asia Pacific, a process of democratisation of science Simon C. Lin EGEE 08, Istanbul

30 Sustainability Issues in Asia Pacific International collaboration is a necessary driver for many countries Participation to WLCG/EGEE/EGI, EUAsiaGrid and OSG are good examples. successful regional coordination model established - APROC and APGI NGI model is not feasible for most Asia countries Long-term funding Advantage: National Grid/e-Science/NII program available: AU, JP, KR, CN, TH, MY, IN and TW Challenge: much relied on local scientific communities Opportunity: International collaborations create much funding opportunities User Community Engagement is really essential Collaboration among research groups and between ICT infrastructure and user community sometimes are not that easy. Regional common application should drive the growth, eg, disaster mitigation, environmental changes, etc. Technology Development Buildup and join the global e-infrastructure by community-based collaboration is the best formula. Local requirements and technology development could be fulfilled within the collaboration framework. Convergence: better to have common platform and interoperable infrastructure locally in each country Interoperation different collaboration might have different flavor infrastructure and technology standard-based interoperation is always necessary, eg, OGF. Security risks Users might need more sophisticated access right management system sometimes. Federated identity management is required to facilitate the convergence of infrastructure, technology and service. Reliability, efficiency and evolution of e-infrastructure: APROC again is the best model for now. 30