Speaker
P. Nevski
(BROOKHAVEN NATIONAL LABORATORY)
Description
The ATLAS detector is a sophisticated multi-purpose detector with
over 10 million electronics channels designed to study high-pT
physics at LHC. Due to their high multiplicity, reaching almost
hundred thousand particles per event, heavy ion collisions pose a
formidable computational challenge. A set of tools have been created
to realistically simulate and fully reconstruct the most difficult
case of central Pb-Pb collisions (impact parameter < 1 fm) in the
ATLAS detector.
A number of issues concerning extensive memory management, CPU
versus memory optimization, tradeoff between data volume and physics
analysis capacity have been formulated and solved. As a result we
are able to predict and optimise the physics performance of the
experiment and its sub-systems. We will describe the optimal
dataflow organization and solutions which allowed flexible system
tuning during the massive simulated data production and the analysis
of tens of thousand of multi-megabyte events.
Primary author
P. Nevski
(BROOKHAVEN NATIONAL LABORATORY)
