Abstracts and Program

PROOF 2007 Workshop

Parallel Distributed Data Analysis for the LHC

29 and 30 November, 2007 at CERN

As the date of first LHC collision is steadily approaching, preparing the end-user data analysis environment is becoming more urgent.

Over the past few years, the PROOF system has been developed to provide a seamlesss extension of the single-user ROOT session to large clusters. Using PROOF the analysis is done in parallel on the cluster, thereby reducing the analysis time and increasing the amount of data that can be processed.

In the meanwhile the computing landscape is changing with the arrival of dual and quad-core CPU's and desktop machines with 32 or more cores are on the horizon. To harness all these cores, programs must be multi-threaded and parallelized. PROOF already runs efficient on 8 core machines and is designed to scale to many more.

This small workshop is organized to bring together people who plan to run PROOF as Central Analysis Facility (CAF) on hundreds of nodes and TB's of disk, or who plan to run it on medium and small departmental clusters (Tier 2 and 3) for physics workgroup analysis.

The main workshop topics are data set management, user and resource scheduling, best hardware to run PROOF on, cluster monitoring and management, experiment analysis models and frameworks, missing features and user feedback.

Fons Rademakers
Last update 30/10/2007 by FR