Presentation is loading. Please wait.

Presentation is loading. Please wait.

Acronyms GAS - Grid Acronym Soup, LCG - LHC Computing Project EGEE - Enabling Grids for E-sciencE.

Similar presentations


Presentation on theme: "Acronyms GAS - Grid Acronym Soup, LCG - LHC Computing Project EGEE - Enabling Grids for E-sciencE."— Presentation transcript:

1 Acronyms GAS - Grid Acronym Soup, http://www.gridpp.ac.uk/gas/http://www.gridpp.ac.uk/gas/ LCG - LHC Computing Project EGEE - Enabling Grids for E-sciencE gLite - Lightweight Middleware for Grid Computing (EGEE) OSG - Open Science Grid PANDA - Production and Distributed Analysis (USATLAS & OSG) Nordugrid - Nordic Grid ARC - Advanced Resource Connector (Nordugrid) WMS - Workload Management System GANGA - Gaudi/Athena and Grid Alliance pathena – parallel Athena DDM - Distributed Data Management DPM – Disk Pool Manager dCache – disk pool management system CASTOR - CERN Advanced STORage manager SRM - Storage Resource Manager FTS - File Transfer System AOD - Analysis Object Data TAG - Event Selection Data

2 2007 is a busy year … Many Distributed Analysis Tutorials – Edinburgh Milano – Lyon – Munich – Toronto – Bergen – Valencia – CERN – Stanford - Sharply rising number of grid users Increasing demand for user support

3 Overview The Grid Infrastructure – EGEE/gLite, OSG/PANDA, Nordugrid/ARC – Data Distribution – Data Access The Distributed Analysis Activity – pathena/PANDA – GANGA/EGEE & Nordugrid – GANGA/PANDA – Dashboard Conclusions

4 Three Grids EGEE – gLite Middleware OSG in the US – PANDA Nordugrid in Scandinavia – ARC Middleware

5 Data Distribution More manpower for the developer team New Data Placement Coordinator New DDM version 0.3 Disk space has been reviewed Eagerly awaiting – SRM 2.2 – New version of FTS Aim: 100 % of data on all T1’s and then further copies on T2’s Today: On 5 T1’s we have 80 to 92 % Problems are related to disk space and to changes to new storage systems

6 Data Access We need direct access to the data – TAG analysis – Faster IO using Athena v13 (2 to 10 MB/sec) Together with the LCG Application Area – Solve issues with the access to various Storage Systems (DPM, dCache, Castor,…) We are also investigating xrootd technology

7 pathena/PANDA Lightweight submission client to PANDA Web pages for monitoring pathena AnalysisSkeleton_jobOptions.py --inDS trig1_misal1_mc12.005200.T1_McAtNlo_Jimmy.recon.AOD.v12000601 --outDS user.DietrichLiko.test.01

8 PANDA since Feb 2006 Close to 200 users since February 2006, about 38 users last month Large number of jobs BNL is main site

9 GANGA/EGEE + Nordugrid GANGA job submission tool – ATLAS and LHCb GUI ganga athena AnalysisSkeleton_jobOptions.py --inDS trig1_misal1_mc12.005200.T1_McAtNlo_Jimmy.recon.AOD.v12000601 --outDS user.DietrichLiko.test.01 --lcg –ce ce-fzk.gridka.de:2119/jobmanager-pbspro-atlasS

10 GANGA since Feb. 2007 ATLAS Users: In total about 440, about 80 users per week

11 GANGA/PANDA GANGA plug-in has been developed to interface GANGA with PANDA pathena is an essential part of the plug-in The same job can be sent now to all ATLAS grid infrastructures – First prototype is available – Working on support for all options ….

12 ARDA Dashboard Provides an overview of all ATLAS activities – LCG System Monitoring WG – PANDA – Nordugrid

13 Conclusions Usage of the Grid for Distributed Analysis has been steadily rising in ATLAS – Many Distributed Analysis tutorials – Many new users – Increasing demand for user support All ATLAS grids are being used for analysis – OSG, EGEE, Nordugrid Good collaboration between developers of the analysis tools

14 gLite WMS A nearly final version is available at CERN – We expect it to be fully certified soon Bulksubmission – Without Condor DAG’s Improved sandbox handling Support for new CE etc. WMS site efficiency at IN2P3 for different experiments CMS/CrabLHCb/DIRAC Alice/Alien ATLAS/GANGA

15 PANDA on LCG PANDA is available as test installation on several LCG sites – Lyon + French cloud – Several other sites The issue is also related to ATLAS Production Ongoing discussion between the involved parties

16 User evaluations At March ATLAS software week At a recent PAT forum Notes are in preparations – CSC Note of SUSY team

17 17 Conclusions Both schemes are user-friendly. Differences between the GANGA and PanDA exist, but user can switch fairly easy from one to the other. Full user control over the job configuration Performance wise, analysis over large datasets can be performed in few hours. Many thanks to Caroline Collard, Stathes Paganis, Mark Hodgkinson, Dimitris Varouchas, Nurcan Ozturk and the H/A  μμ group for their feedback. Having the experience of running alone a small computer cluster and using the old-fashioned distributed analysis scheme, my conclusion is that using PanDA and GANGA is easier and less time consuming. Moreover, all feed-back received was very positive! (although there are still a few missing pieces) Concerns: 1) Data replication uniformity over the sites. 2) Integrity of data-files. Recovery of crashed jobs due to corrupted files not easy. (currently, user manually isolate the problematic files and re-submit the job in the old way) 3) Distinction between LCG and OSG not very productive. (situation is improving though, eg PanDA now runs on site ANALY_LYON) 4) Provide stable environment for users during development. 5) Maintain good performance with increasing number of users.


Download ppt "Acronyms GAS - Grid Acronym Soup, LCG - LHC Computing Project EGEE - Enabling Grids for E-sciencE."

Similar presentations


Ads by Google