Biomedicineand Life Sciences II Sijin Qian

25 %
75 %
Information about Biomedicineand Life Sciences II Sijin Qian

Published on November 1, 2007

Author: Herminia


Grid Computing Program at Peking University in EUChinaGRID Project:  Grid Computing Program at Peking University in EUChinaGRID Project Outline:  Outline EUChinaGRID project and PKU group Grid infrastructure at PKU (School of Physics) WP4 (for Grid application) activities at PKU Biology subgroup: Protein structure analysis Physics subgroup: CMS Monte-Carlo simulation and physics analysis Main problems and solutions Networking Software installation at Grid sites Summary EUChinaGRID Project 欧中网格项目 (More details will be presented by Dr. Giuseppe ANDRONICO tomorrow):  EUChinaGRID Project 欧中网格项目 (More details will be presented by Dr. Giuseppe ANDRONICO tomorrow) Project Banner:  Project Banner Interconnection and Interoperability of Grids between Europe and China Timescale & Budget:  Timescale & Budget The official start of the project: 1st January 2006. Duration: 24 Months EU Contribution: 1,299,998 €. A total 495 Person Months (325 Funded) of effort Partners:  Partners Third Parties:  Third Parties Targets of the Project:  Targets of the Project To foster the creation of a intercontinental eScience community Training people Supporting existing and new applications To support interoperable infrastructure for grid operations between Europe (EGEE) and China (CNGRID) WPs (Working Packages) :  WPs (Working Packages) Work Breakdown Structures:  Work Breakdown Structures Collaborative tools:  Collaborative tools Project Web Sites:  Project Web Sites and (English) (Chinese中文) Infrastructure 基础设施:  Infrastructure 基础设施 What we have already done:  RB (Resource Broker) + BDII (Berkely Database Information Index) at CNAF (Italy) VOMS at CNAF GridIce(Grid sites monitoring)at CNAF Sites linked: Roma 3 (Italy) CNAF (Italy) Catania (Italy) Athens (Greece) 3 sites in Beijing (CNIC, IHEP and PKU) What we have already done Sites Map:  Sites Map Sites Monitoring:  Sites Monitoring BEIJING - PKU Training Program:  April 3-7, 2006 in Beijing, China (done) April 18-21, 2006 in Rome, Italy (done) June 12-16, 2006 at IHEP + Project’s 1st Workshop in Beijing, China (done) September 15-22, 2006 in Rome, Italy + Project’s 1st Conference (done) November 25-26, 2006 at Peking University (done). All Chinese tutors in first time. April 16-20, 2007 at CNIC, Beijing, China Training Program Peking University in EUChinaGRID Project:  Peking University in EUChinaGRID Project Subgroups & Personnel:  Subgroups & Personnel Biological Research – Protein structure study with NMR (led by Prof. B. XIA,夏滨) C. JIN, Y. FENG, W. GONG, X. GUO, T. WANG. To participate in WP4 (4.3) High Energy Physics Research – CMS experiment on LHC at CERN (led by Prof. S. QIAN,钱思进) Z. YANG, L. ZHAO, D. MU, S. ZHU, K. KANG To participate in WP4 (4.1) and WP3 Also, both groups are working in WP5 Biology Group:  Biology Group Beijing NuclearMagneticResonance Center:  Beijing NuclearMagneticResonance Center Sponsored by Ministry of Science and Technology, Ministry of Education, Chinese Academy of Science, Chinese Academy of Military Medical Sciences, Managed by Peking University. National NMR facility established on Nov. 4th, 2002 For research and training in bio-molecular NMR studies We need to use computer for processing and analyzing NMR data, for solution structure calculation, and for molecular dynamic simulation. Slide22:  Key method for obtaining high resolution structure -----in addition to X-ray Structure Physiological temperature and condition -----closer to native functional state Time consuming for structure calculation -----multiple structures and multiple rounds NMR Spectroscopy Slide23:  NMR Structure Determination Slide24:  From Constraints to Structure Restrained molecular dynamics and simulated annealing Slide25:  V = Eempirical + Eeffective with: Eeffective = ENOE + Etorsion and Eempirical = Ebond + Eangle + Edihedral + Evdw + Eelectr Empirical energy contains all information about the primary structure of the protein and also data about topology and bonds in proteins in general. Empirical energy are from experimental data. Force Field Slide26:  Energy Minimization Slide27:  Structure Calculation and Refinement Normally, 200 structures/round, > 30 rounds. Slide28:  Recent Structures Analysis Software:  Analysis Software Protein structure analysis software: Amber. Licenses are needed to be granted on all computers involved. University Rome III has procured the license and is testing it, hopefully it can be available for use in near future. PKU-Biology Computing Need:  PKU-Biology Computing Need By using the Intel 2.4 GHz Xeon CPU Each structure needs 4 hours Each time to compute 200 structures Each protein needs to be computed for 10 times Totally 10 proteins to be analyzed  ~ 80,000 hours (> 9 years) CPU time > 1TB storage space Physics Group:  Physics Group Physics Data Analysis for CMS Experiment:  Physics Data Analysis for CMS Experiment CMS group in the Physics School of Peking University has started to use Grid tools to analyze physics data of CMS experiments on LHC at CERN since 9/2005 Huge amount of Monte-Carlo data (from now on) and real data (collected from the end of 2007) shall await for us to analyze 27 km circumference LHC completion date: 2007.11 LHCComputingGrid Model:  LHCComputingGrid Model LCG Architecture at PKU :  LCG Architecture at PKU Installed at PKU (UI) (SE) (CE) (WN) (SE) Installed at PKU (UI) (CE) @IHEP Working History:  Working History Single J/y  m+m- generation (without background) and reconstruction by using local computers in 6/2005 Single J/y study with min-biased background in 7/2005 Analyzed 500 B0s  J/y + f events from a DST (Data Summary Tapes) at CERN in 8/2005 Analyzed nearly 200,000 B0s events from a DST stored in Italy by using Computing Grid tools from 9/2005 and going on Preparing the massive (> 2 millions J/y events) Monte-Carlo simulation Procedure of Grid Application:  Procedure of Grid Application The latest procedure via the IHEP LCG Tier-2 facility: PKU’s UI gets the results from submit the jobs IHEP’s RB run the jobs, send the jobs to CE return the results to IHEP’s RB give the jobs to WN UI (User Interface)@PKU, China RB (Resource Broker)@IHEP, China CE (Computing Element)@CNAF, Italy WN (Work Nodes)@CNAF, Italy Sample Result:  Sample Result J/psi reconstruction efficiency as a function of PT (both muons’ |eta|<=2.4) J/y reconstruction efficiency in CMS experiment First CMS Analysis Note by Peking Univ. Group:  First CMS Analysis Note by Peking Univ. Group PKU-Physics Computing Need:  PKU-Physics Computing Need In 2007, we would wish to generate > 2 million events each for prompt J/Psi and Upsilon + 40% of background events For each 1 million events, it needs about 24,000 hours (or 1000 days) of CPU time (for one P4 Xeon 1.5GHz computer), and about 1.1 TB of storage space. In result, we would need ~5600 days (i.e. ~ 18 years) of CPU time & ~6 TB of storage space Summary of WP3 & WP4 Activities at PKU :  Summary of WP3 & WP4 Activities at PKU Established a LCG (LHC Computing Grid) Tier-3 site for getting access to the LCG system; Used the above system to have analysed a large MC dataset stored at CNAF in Italy, and have produced some analysis results; Provided configuration files for CMS collaboration in order to generate >2 million prompt J/y events; Installed the CMSSW on EUChinaGrid system (Catania site); Preparing the protein structure analysis in Biology group; Has estimated the computer and storage resources needed to handle the millions of events for Physics group and to analysis the protein structure in Biology group. Main Problems:  Main Problems Availability of biological software (Amber) Licensing Stability of CMS software (CMSSW) the suitable J/y event generator is still being tested by CMS collaboration before to be put in production HLT (High Level Trigger) software Networking Bandwidth (international traffic is charged by bits) University policy (3 levels of gateway) Networking in PKU:  Networking in PKU 3 levels of gateway Campus network: no charge, only within campus Domestic gateway: minor monthly charge, unlimited traffic International gateways: Monthly package -- 90 Yuan/month, unlimited traffic, but disconnected every few hours if no activities Server gateway -- no interruption, but charged by bits Solutions:  Solutions Use the domestic gateway to connect to IHEP via VPN (Virtual Private Network), then to reach the world through the IHEP’s trunk line. Applied and installed the CERNET’s special link to TEIN2. The special cabling was done in 1/2007. No charge by bits No periodical interruption. Network Topology Map:  Network Topology Map The improved route (TEIN2): will upgrade to 2.5 Gbps The backup route Summary:  Summary PKU group has set up a very basic Grid site for getting access to the LCG system and for preparing the massive biological protein structure analysis. By using this system, we have engaged in some CMS physics study and got some encouraging results. Some long standing problems of networking have been finally solved with the TEIN2 connection. Much more works are to be done, we must start the protein structure analysis as soon as the software licence is granted; be fully prepared for the CMS data analysis when LHC’s first proton beam collision at the end of 2007. Slide46:  Thank you (謝謝)!

Add a comment

Related presentations