15
Belle MC Production Belle MC Production on Grid on Grid 2 2 nd nd Open Meeting of the SuperKEKB Collaborat Open Meeting of the SuperKEKB Collaborat ion ion Soft/Comp session Soft/Comp session 17 March, 2009 17 March, 2009 Hideyuki Nakazawa Hideyuki Nakazawa National Central University National Central University

Belle MC Production on Grid 2 nd Open Meeting of the SuperKEKB Collaboration Soft/Comp session 17 March, 2009 Hideyuki Nakazawa National Central University

Embed Size (px)

Citation preview

Belle MC ProductionBelle MC Production on Grid on Grid

22ndnd Open Meeting of the SuperKEKB Collaboration Open Meeting of the SuperKEKB CollaborationSoft/Comp sessionSoft/Comp session17 March, 200917 March, 2009Hideyuki NakazawaHideyuki NakazawaNational Central UniversityNational Central University

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 22

ContentsContents

Status of MC productionStatus of MC production Belle MC production procedure with GridBelle MC production procedure with Grid GangaGanga

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 33

MC Production StatusMC Production Status

Long down time since last JuneLong down time since last June Just resumed Just resumed New sitesNew sites

KISTI, Melbourne, Ljubljana, KarlsruheKISTI, Melbourne, Ljubljana, Karlsruhe

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 44

Belle VO sitesBelle VO sitesInstituteInstitute CPUCPU StatusStatus

KISTIKISTI 6060 newnew

MerbourneMerbourne 6060 newnew

KarlsruheKarlsruhe 400400 newnew

LjubljanaLjubljana 8080 newnew

ASGCASGC 120120 Stop due to fire incidentStop due to fire incident

KEK01KEK01 1212 Unavailable due to old OSUnavailable due to old OS

KEK02KEK02 3232 Under upgradingUnder upgrading

KrakowKrakow 9090

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 55

One Week StatisticsOne Week Statistics

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 66

ProblemsProblems Storage, HSM busyStorage, HSM busy

Stays inside B system and shared with Belle users Stays inside B system and shared with Belle users Access to HSM sometimes concentrate on one of Access to HSM sometimes concentrate on one of

four serverfour server

Job processing time exceed limitsJob processing time exceed limits

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 77

Belle MC ProcedureBelle MC Procedure 1. Get Assignment from Web1. Get Assignment from Web

4 physics modes (charged, mixed, charm, uds)4 physics modes (charged, mixed, charm, uds) 1 slot = 100 run range1 slot = 100 run range

Same number of events with experiment dataSame number of events with experiment data 10 set (streams)10 set (streams)

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 88

Belle MC ProcedureBelle MC Procedure 2. Receive a mail 2. Receive a mail

Gen and Bkg files: copy to HSM (in B

computer system) and register to SRB “Sregister hoge.pgen srbdir”

Copy script.tgz to UI

SRB(Storage Resource Broker)Accessible to/from GridUI(User Interface)Server you work for Grid

Generator, Bkg, sh script files are provided by MC manager

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 99

Belle MC ProcedureBelle MC Procedure 3. Submit jobs3. Submit jobs

1. Create a directory for the slot (prepare-job.rb)2. Modify sh script and create JDL files (basf2grid.sh)

Add “copy inputs from SRB” and “copy outputs to SRB”3. Create a sh script to submit (mksubmit.rb)

“edg-job-submit … job.jdl”4. Submit (submit)

Executable=Job.shStdout=log.outStderr=log.err

JDL

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1010

Belle MC ProcedureBelle MC Procedure 4. Monitor running jobs4. Monitor running jobs

Summary CPU

Elapse time

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1111

Belle MC ProcedureBelle MC Procedure 5. Retrieve finished job5. Retrieve finished job

•Retrieve log files of finished jobs•Checking the job is correctly completed

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1212

GangaGanga Ganga is an easy-to-use Ganga is an easy-to-use frontend for job definition and managemfrontend for job definition and managem

entent, implemented in , implemented in PythonPython. It has been developed to meet the ne. It has been developed to meet the needs of the eds of the ATLASATLAS and and LHCbLHCb for a Grid user interface, and include for a Grid user interface, and includes built-in s built-in support for configuring and running applications based osupport for configuring and running applications based on the n the GaudiGaudi / / Athena Athena framework framework common to the two experiments. common to the two experiments. Ganga allows trivial switching between testing on a local batch syGanga allows trivial switching between testing on a local batch system and large-scale processing on Grid resources. stem and large-scale processing on Grid resources.

http://ganga.web.cern.ch/ganga/index.php

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1313

Ganga “Hello World” Ganga “Hello World”

Configure with Job()Configure with Job() DirectoriesDirectories JDL fileJDL file

Backend = localBackend = localConfiguration

Submit

Status

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1414

Ganga MC ProductionGanga MC Production

→ edg-job-submit

Backend=LCG

Mar17, 2009Mar17, 2009 Belle MC with GridBelle MC with Grid 1515

SummarySummary

MC Production resumed !!MC Production resumed !! New sites working correctly with large resNew sites working correctly with large res

oucesouces Configuration and Job submission tool “GConfiguration and Job submission tool “G

anga” tried. anga” tried. Easy to see what we’re doingEasy to see what we’re doing Good to unified interfaceGood to unified interface No convenience about functionalityNo convenience about functionality

We must write script for scripts, anyway.