Speaker
T.M. Steinbeck
(KIRCHHOFF INSTITUTE OF PHYSICS, RUPRECHT-KARLS-UNIVERSITY HEIDELBERG, for the Alice Collaboration)
Description
The Alice High Level Trigger (HLT) cluster is foreseen to consist of
400 to 500 dual SMP PCs at the start-up of the experiment. The
software running on these PCs will consist of components
communicating via a defined interface, allowing flexible software
configurations. During Alice's operation the HLT has to be
continuously active to avoid detector dead time. To ensure that the
several hundred software components, distributed throughout the
cluster, operate and interact properly, a control software was
written that is presented here. It was designed to run distributed
over the cluster and to support control program hierarchies.
Distributed operation avoids central performance bottlenecks and
single-points-of-failures. The last point is of particular
importance, as each of the commodity type PCs in the HLT cluster
cannot be relied upon to operate continously. Control hierarchies in
turn are relevant for scalability over the required number of nodes.
The software makes use of existing and widely used technologies:
Configurations of programs to be controlled are saved in XML, while
Python is used as a scripting language and to specify actions
to execute. Interface libraries are used to access the controlled
components, presenting a uniform interface to the control program.
Using these mechanisms the control software remains generic and can
be used for other purposes as well. It is being used for HLT data
challenges in Heidelberg and is planned for use during upcoming beam
tests.
Primary authors
H. Tilsner
(KIRCHHOFF INSTITUTE OF PHYSICS, RUPRECHT-KARLS-UNIVERSITY HEIDELBERG, for the Alice Collaboration)
T.M. Steinbeck
(KIRCHHOFF INSTITUTE OF PHYSICS, RUPRECHT-KARLS-UNIVERSITY HEIDELBERG, for the Alice Collaboration)
V. Lindenstruth
(KIRCHHOFF INSTITUTE OF PHYSICS, RUPRECHT-KARLS-UNIVERSITY HEIDELBERG, for the Alice Collaboration)
