o. stézowski ipn lyon agata week 15-19 september 2003 legnaro data analysis – team #3 root as a...

20
O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro Data Analysis – Team #3 ROOT as a framework for ROOT as a framework for AGATA AGATA

Post on 19-Dec-2015

219 views

Category:

Documents


0 download

TRANSCRIPT

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Data Analysis – Team #3Data Analysis – Team #3

ROOT as a framework for AGATAROOT as a framework for AGATA

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Data Storage Online / offline monitoring and analysis

What is behind « T3 » ?What is behind « T3 » ?

Everything needed once

the beam is on the target

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

OutlineOutline

What is currently done What should be done Why ROOT can be used Conclusions

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

What is currently done ?What is currently done ?

Data storageDLT, disk, BLUE, StasbourgDB, Radware

AnalysisMIDAS, Radware, paw, others …

LanguageFortran, C, C++, MIDAS, Java, others …

Is it the best way to work for AGATA ?Is it the best way to work for AGATA ?

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

« Framework approach »« Framework approach »

AGATA Soft

Requirements: Portable / evolutive / well designed

User AUser B

User C

Starting from scratch or from an existing framework ?Starting from scratch or from an existing framework ?

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

“It became time to re-think our approach to large scale data analysis and simulation and at the same time we had to profit from the progress made in computer science over the past 15 to 20 years. Especially in the area of Object-Oriented design and development. Thus was born ROOT.”

Why ROOT ?Why ROOT ?

From PAW to ROOT …

http://root.cern.ch

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

OverviewOverview

• Language C++, Open source• First version in 1997 updated version every six months• Large community involved (4 workshops)• Portable on many platforms unix, linux, windows, OS X CC, gcc, icc

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Libraries of objects for physicistsLibraries of objects for physicists

• More than 500 classes already defined ex: histograms, functions, graphs, minimization classes…

• Many methods (TH1 ~210)

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

ROOT interpreter ROOT interpreter

C/C++ interpreter Ex :

root > TF3 f(‘name’,’sin(x*x+y*y+z*z-36)’,-2,2, ,-2,2 ,-2,2)root > f.Eval(1,0,1)root > f.Draw()

Complex macros & functions There is a way to compile macros There is a way to extend the root dictionary

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Object persistence: ROOT filesObject persistence: ROOT files

Take into account the object evolution – machine independent

Successive updates

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Many graphical featuresMany graphical featuresWindows (canvas/pad) to display any ROOT object

• Easy to divide a window• Windows saved in: ROOT file, postscript, svg, gif

Rich GUI

• Move, expand, delete

• Access to some methods

• Way to add new methods

The ROOT geometry package is a tool designed for building, browsing, tracking and visualizing a detector geometry.

The goal is to be able to use the same geometry for several purposes, such as tracking reconstruction or visualization

Concerning the tracking:

ROOT does not track itself, it proposes an interface. The real work is done by what is plugged at the running time (GEANT 3 or 4)

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Event displayEvent display

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Others Others

• Interface to the operating system• Generation of random numbers (MC) ex: random number from an histogram• Many ways to fit• Base classes to built new GUI• Thread support• Map files • …..

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

TTree: special object to store eventsTTree: special object to store events

E U R O B A L L G a m m aM , E 1 , .. .E n

B G O B A LLH , K

D IA M A N TM , Id , E

E ve n tE U R O B A L L IV + D IA M A N T

• Each branch can be splitted• Methods to apply cuts• Treeviewer to display distributions• Mechanism to chain trees

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Proof and GRIDProof and GRID

Proof Parallel ROOT Facility

proof 1

ROOT is part of the LCG (LHC Computing Grid) projects

proof 2 proof n

Master proof ROOTROOT

request

out/obj

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Data management - What is needed ?Data management - What is needed ?

1 seconde 1 minute 1 hour 1 day

4 105 7 103 116 5

How many parallel processes are neededHow many parallel processes are neededto go through all to go through all the data setthe data set

GRID !!GRID !!

40 TB - Data access time 100 MB/sec40 TB - Data access time 100 MB/sec

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

What is done in our fieldWhat is done in our field

Sofware solutions developed to speed up the gating process BLUE, StrasbourgDB, Radware

• Data ~50 GB + Table ~50 GB• 10 stations running at the same time• GRETA liked simulated events (simple ones)• fold gating ~ 6-7

Radware next generation developed for GRETA

50 GB [50 MB/s], reduction by a factor 1050 GB [50 MB/s], reduction by a factor 1033gated spectrum in seconds !!gated spectrum in seconds !!

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Data management - What is needed ?Data management - What is needed ?

1 seconde 1 minute 1 hour 1 day

4 105 7 103 116 5

400 7 1 1

GRID ??GRID ??

40 TB - Data access time 100 MB/sec40 TB - Data access time 100 MB/sec

How many parallel processes are needed How many parallel processes are needed to go through all to go through all the interesting eventsthe interesting events

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

Questions to be answeredQuestions to be answered

• « Cluster of Universities » for each experiment• AGATA center• GRID

Data storage?Disk (cheap, fast), tapes

What structure for the parallel processing?

Avoid too much data transfert between computers !!Avoid too much data transfert between computers !!

• It costs « probably » too much time• It becomes difficult to handle

O. Stézowski IPN Lyon AGATA Week 15-19 September 2003 Legnaro

ConclusionsConclusions

ROOT: solid framework for AGATA Specific developments for Nuclear Physics

Our objects: NuclearLevel, Gamma …Our methods: Gating, DCO, …

Any suggestions are welcome !Meeting tomorrow & AGATA-ANALYSIS

EX Database (Oracle, MySQL) of level scheme