h. kornmayer magic-grid status report egaap meeting, athens, 21th april 2005 - 1 enablinggrids for...

H. Kornmayer MAGIC-GRID Status report EGAAP meeting, Athens, 21th April 2005 - 1

EnablingGridsfor E-siencEGeneric Application:

Monte Carlo Production

for the MAGIC telescope

MAGIC Grid Status report

Harald Kornmayer, Alberto Forti, Toni Coarasa

Mirco Mazzucato, Manuel DelfinoForschungszentrum Karlsruhe; University of Udine;

Max-Planck-Institut für Physik, München; CNAF/INFN Bologna; PIC, Bologna

on behalf of the MAGIC collaboration


EnablingGridsfor E-siencE

Outline• Introduction

• What kind of MAGIC?

• DevelopmentUse Cases RequirementsWorkflows MetaData

• StatusHardware SoftwareDeployment

• ExperienceFirst Month on production is overGrid-added value



Introduction: The MAGIC Telescope• Ground based

Air Cerenkov Telescope• Gamma ray: 30 GeV - TeV• LaPalma, Canary Islands

(28° North, 18° West)• 17 m diameter• operation since autumn 2003

(still in commissioning)• Collaborators: IFAE Barcelona, UAB Barcelona,

Humboldt U. Berlin, UC Davis, U. Lodz, UC Madrid, MPI München, INFN / U. Padova, U. Potchefstrom, INFN / U. Siena, Tuorla Observatory, INFN / U. Udine, U. Würzburg, Yerevan Physics Inst., ETH Zürich

Physics Goals: Origin of VHE Gamma raysActive Galactic NucleiSupernova RemnantsUnidentified EGRET sourcesGamma Ray Burst



~ 10 kmParticleshower

Ground based γ-ray astronomy

~ 1o

Ch

eren

kov

ligh

t

~ 120 m

Gammaray

GLAST (~ 1 m2)

Cherenkov light Image of particle shower in telescope camera

reconstruct: arrival direction, energyreject hadron background



MAGIC – Hadron rejection

• Based on extensive Monte Carlo Simulation

• air shower simulation program CORSIKA

• Simulation of hadronic background is very CPU consuming

• to simulate the background of one night, 70 CPUs (P4 2GHz) needs to

run 19200 days

• to simulate the gamma events of one night for a Crab like source takes

288 days.

• At higher energies (> 70 GeV) observations are possible already by

On-Off method (This reduces the On-time by a factor of two)

• Lowering the threshold of the MAGIC telescope requires new

methods based on Monte Carlo Simulations



Developments - Requirements

• MAGIC needs a lot of CPU to simulate the hadronic background to

explore the energy range 10 GeV – 100 GeV

• MAGIC needs a coordinated effort for the MonteCarlo production

• MAGIC needs an easy accessible system

(Where are the data from run_1002 and run_1003?)

• MAGIC needs an scalable system (as MAGIC II will come 2007)

• MAGIC needs the possiblity to access data from other experiments

(HESS, Vertias, GLAST, PLANCK(?)) for multiwavelength campaigns



Development - Team• The proposal was written in

March 2004• Support for the proposal from

– CNAF (M. Mazzucato)– PIC (M. Delfino)– FZK – MAGIC

• Organisation– Contact Person

• Harald Kornmayer– VO Manager

• Harald Kornmayer– Gridifying developers

• Alberto Forti• Harald Kornmayer• Toni Coarasa



Development – MC Workflow

I need 1.5 million hadronicshowers with Energy E, Direction (theta, phi), ... As background sample for observation of „Crab nebula“

Run Magic MonteCarlo Simulation and register output data




Run Magic Monte Carlo Simulation (MMCS) and register output data

Simulate the Telescope Geometry with the reflector program for all interesting MMCS files and register output data

Simulate the Starlight Background for a given position in the sky and register output data

Simulate the response of the MAGIC camera for all interesting reflector files and register output data

Merge the shower simulation and the StarLight simulation and produce a MonteCarlo data sample



• Huge number of runs with different parametersWe want to record what we did!

MetaData needed!

MMCS – Meta data

Model• for MMCS

• Store run parameters• Monitor Grid Jobs• Manage MMCS data

• for reflector• prototype designed

• for camera• need to be done!



Status - Hardware

• History:– The Hague, Nov. 2004:

• The MMCS prototype was developed on the CrossGrid testbed

• First test on GILDA

• Migration to real EGEE infrastructure needed

• Now: – Running on real EGEE infrastructure (april 2005)

• VO server at Nihkef/SARA

• BDII and Resource Broker at CNAF

• Replica services at PIC

• Resources (whereever available! come to that later!)

bash-2.05a$ lcg-infosites --vo magic allLRC endpoint for magic: http://rlsmagic.pic.es:8080/magic/edg-local-replica-catalog/services/edg-local-replica-catalogRMC endpoint for magic: http://rlsmagic.pic.es:8080/magic/edg-replica-metadata-catalog/services/edg-replica-metadata-catalogLRC endpoint for magic: http://rlsmagic.pic.es:8080/magic/edg-local-replica-catalog/services/edg-local-replica-catalogRMC endpoint for magic: http://rlsmagic.pic.es:8080/magic/edg-replica-metadata-catalog/services/edg-replica-metadata-catalog

****************************************************************These are the related data for magic: (in terms of CPUs)****************************************************************

#CPU Free Total Jobs Running Waiting ComputingElement---------------------------------------------------------- 20 20 0 0 0 ce01.pic.es:2119/jobmanager-torque-magic 168 4 1 1 0 lcgce02.ifae.es:2119/jobmanager-lcgpbs-magic 7 6 0 0 0 ce001.grid.bas.bg:2119/jobmanager-lcgpbs-magic 82 63 6 6 0 gridba2.ba.infn.it:2119/jobmanager-lcgpbs-long 82 63 3 3 0 gridba2.ba.infn.it:2119/jobmanager-lcgpbs-short1280 5 0 0 0 lcg-gridka-ce.fzk.de:2119/jobmanager-pbspro-long1280 5 0 0 0 lcg-gridka-ce.fzk.de:2119/jobmanager-pbspro-short 50 29 1 1 0 grid012.ct.infn.it:2119/jobmanager-lcgpbs-long 50 29 0 0 0 grid012.ct.infn.it:2119/jobmanager-lcgpbs-short 82 63 8 8 0 gridba2.ba.infn.it:2119/jobmanager-lcgpbs-infinite1280 5 0 0 0 lcg-gridka-ce.fzk.de:2119/jobmanager-pbspro-default 582 413 0 0 0 ce01-lcg.cr.cnaf.infn.it:2119/jobmanager-lcglsf-magic1280 5 0 0 0 lcg-gridka-ce.fzk.de:2119/jobmanager-pbspro-extralong 50 29 24 20 4 grid012.ct.infn.it:2119/jobmanager-lcgpbs-infinite 8 6 1 1 0 gridit-ce-001.cnaf.infn.it:2119/jobmanager-lcgpbs-long 8 6 0 0 0 gridit-ce-001.cnaf.infn.it:2119/jobmanager-lcgpbs-short 8 6 1 1 0 gridit-ce-001.cnaf.infn.it:2119/jobmanager-lcgpbs-infinite----------------------------------------------------------

The total values are:----------------------------------------------------------4889 648 45 41 4

**************************************************************These are the related data for magic: (in terms of SE)**************************************************************

Access point Avail Space(Kb) Used Space(Kb) Files SEs---------------------------------------------------------------------------------------------------------------gsiftp://castorgrid.ifae.es/castor/ifae.es/lcg/magic/ 1000000000000 0 0 castorgrid.ifae.esgsiftp://castorgrid.pic.es/castor/ifae.es/lcg/magic/ 1000000000000 0 0 castorgrid.pic.esgsiftp://castorgrid.ifae.es/castor/ifae.es/lcg/magic/ 53020932 0 0 castorgrid.ifae.esgsiftp://se001.grid.bas.bg/storage/magic/ 354813480 30087 4 se001.grid.bas.bggsiftp://gridba6.ba.infn.it/flatfiles/SE00/magic/ 1605498664 0 0 gridba6.ba.infn.itgsiftp://grid007g.cnaf.infn.it/flatfiles/SE00/magic/ 1497769448 0 0 grid007g.cnaf.infn.itgsiftp://lcg-gridka-se.fzk.de/grid/fzk.de/..../SE00/magic/ 1605498664 0 0 lcg-gridka-se.fzk.de



Status – Hardware

Nikhef/Sara FZK

BA.INFN.IT

PICCNAF

BAS.BG

BDIIMetaDataBase

UI

VO-Server

UI

RB

CE

CE

CE

UI CE

RLS

SECE

MPI.DE

UI CE

CT.INFN.IT

CE CE

This system is running and used for the first Monte Carlo Production challenge on the Grid (12 MEvts)

First experience

EGEE-1

CE

EGEE-2

CE

EGEE-2

CE

(apr/may 2005)

SESE



Status – Software

• MMCS system version 1.0

• is up and running!

• Reflector prototype

• tested

• Camera

• needs to be done!

• Currently we submit all the software with the jdl

no software installation required yet

methods like the GAIT are tested (see Crakow Grid Workshop 2004)

• The submission needs approx. 40 sec

• With a more dedicate application installation procedure, this can be improved!!

• MMCS system version 1.0

• Java based prototype

• Hides the EDG/lcg command from the user

• Access the Meta-database

• Runs on User Interface



Experience – Execution

Data challenge Grid-1• 12M hadron events

• 12000 jobs needed• started march 2005• up to now ~ 4000 jobs

First tests:• with manual GUI submission• Reasons for failure

• Network problems• RB problems• Queue problems

Job successful:Output file

registered at PIC

Diagnostic:• no tools found• complex and time consuming use metadata base, log the failure, resubmit and don‘t care

170/3780 Jobs failed 4.5 % failure



Experience – InputProcOutput

• How to submit a job?

JDL

• JDL should specify

• Get the input form

• InputSandBox

• InputData

• Run the program

• Executable

• Store the output at

• OutputSandBox

• OutputData

Inpu

tDat

a

Out

putD

ataFile on UI

InputSandBox

File on GridInputData

File to UIOutputSandBox

File to GridOutputData

OK!

No file transfer!

OK!

OK!

Answer from experts: write a script that copy the file

a SE to the WN BUT: I don‘t want to implement a WORKAROUND

for basic grid functionality



• The VO is managed with the VO server

• Why do the MAGIC grid users needs to be registered at CERN?

• Clarification of the non-LCG-registration procedure needed!!

• SA1 centers must know about this procedure!!

• VO manager has no means to rejected offered resources

• Only the local sites decide, if they want to support the VO

• There is no certification/test procedure before

• this can produce instability of a running production system!

• The VO manager can not remove them!

• Who is managing the VO?

Experience – VO Management



• Qualitative

• Enable „(e-)scientific“ collaboration between partners

• Enable the cooperation between different experiments

• „Towards a virtual observatory for VHE gamma rays“

(presentation next week)

• Enable the participation on Virtual Observatories

Grid added value

• Quantitative

• Number of temporarily available CPU unreachable

no money, no manpower

• Aim: 1.5 Mio Events in 24 hours

• Manpower for storage maintenance reduced



Technical:

• „missing“ functionality in

middleware

• IPO definition

• Job status information to

computing centric

• VO management

• Set up

• Local mapping

• Central registration

bottlenecks

Technical:

• Tools for error diagnostics not

available

• VO supporting tools

• More Batch systems need

support

• Sun Grid Engine

• Firewall

• The range of open firewall

ports is a problem for new

computing centers



Political:

• …

bottlenecks

Money:

• Near-future availablility

• EGEE-2 resources (?)

• Beyond 2007 (?)

• Far-future Grid operation

models?

• Sustainablility ?

Manpower:

• EGEE Account Manager

Administrative:

• Single point of contact

VO account manager

• Many boards

• EGAAP vs. OAG

• Many documents read

• Where to Send Middleware

Requirements?

• Communication flow inside

EGEE not obvious for new

VOs



EGEE – MAGIC Grid

• From TheHague to Athens• MAGIC Grid on EGEE became real

• First Monte Carlo Production challenge • 12 Mio Events • started mid of March• ongoing

• Evaluation of a Grid based analysis system for MAGIC

• New application domain on the Grid•„Towards a virtual VHE gamma ray observatory“

wait for the future

h. kornmayer magic-grid status report egaap meeting, athens, 21th april 2005 - 1 enablinggrids for...

Documents