Krzysztof Korcyl

Seeking for Computing Resources by the ATLAS HLT System


Abstract

Performance of the ATLAS High Level Trigger (HLT) system strongly
depends on a number of computing nodes where filtering algorithms
can be executed. Current, optimistic estimate soars for more than
3500 CPUs. Vast majority of these nodes will be installed at CERN,
hence the actual version of the HLT software does not foresee
usage of resources located outside of CERN.  However, the remaining
CPUs need to be sought for in remote sites, but also the procedure
for acquiring more computing resources will be necessary when
the ATLAS physics programme will be experimented with and more
CPUs may be needed in a dynamic way.

The live demo presentation shows possible procedure to gain
remote resources from the Crossgrid testbed. It uses current
concept of the HLT software replacing only one component with a
proxy and making the change invisible for the rest of the software.
It is expected that the proposed procedure will become a ground for
more sophisticated studies on using grid resources for real time,
throughput and processing hungry applications.