IO performance of ATLAS data formats

The interplay of the ATLAS persistent event data model and the ROOT based I/O backend was studied in order to improve the read performance and disk size of ATLAS data formats for simulation, reconstruction and data analysis. The enabling of several native ROOT features such as basket ordering and th...

Descripción completa

Detalles Bibliográficos
Autores principales: Schaffer, RD, Bhimji, W, Biscarat, C, Brandt, G, Duckeck, G, van Gemmeren, P, Peters, A, Vukotic, I
Lenguaje:eng
Publicado: 2010
Materias:
Detectors and Experimental Techniques
Acceso en línea:http://cds.cern.ch/record/1299564
Descripción
Sumario:The interplay of the ATLAS persistent event data model and the ROOT based I/O backend was studied in order to improve the read performance and disk size of ATLAS data formats for simulation, reconstruction and data analysis. The enabling of several native ROOT features such as basket ordering and the tree cache has lead to significant improvements in dedicated test setups using local disks, and file servers managed by dCache, DPM and xrootd. After implementation in the ATLAS Athena framework, tests in more realistic environments were done. The functionality of the improvements and results from the performance tests are reported.

Ejemplares similares