Apr 1, 2000
The COMPASS off-line computing
University of Trieste and INFN Trieste
COMPASS is a fixed target experiment at the CERN SPS.
The commissioning of the apparatus will start in May 2000.
The diversified physics program includes several measurements
with both muon and hadron beams, which require high statistics,
and slightly different configurations of the experimental
The experiment will take data for at least 5 years, for about
six months each year. The estimated continuous data acquisition
flux is of the order of 35 MB/s, corresponding to about 300 TB
of RAW data to be stored and processed on the COMPASS Computing
Farm (CCF) each year. The Central Data Recording and the RAW data
processing will be performed in real time on the CCF, consisting
of about 100 dual PCs, and presently under test (see E23).
To guarantee flexibility and maintenability of the off-line software,
COMPASS decided to build a completely new system, using OO techniques,
C++ as programming language and Objectivity/DB to store all the data.
Presently, the software off-line system is in an advanced phase of
construction and test. The goal is to have a first version with all
the foreseen functionalities ready by May 2000, to start tests with
real data during the commissioning of the experiment.
The data structures are well defined and designed; in particular
the Event DB structure is ready since several months.
CORAL (COMPASS Reconstruction and AnaLysis program), written
in C++ with OO techniques, has a modular structure being based
on a framework (ready) providing an analysis class library,
interfaces for the event reconstruction algorithms, insulations
layers for the access to the DBs (HepODBMS) and for all external
packages (in particular ROOT, presently used for histogramming).
Most of the event reconstruction algorithms are ready and have been
fully tested with MC (Geant3) events. Other methods, like event
buffer decoding, are under implementation, in common with on-line
and detector groups.
Most of the difficulties COMPASS had (and has) in preparing the new
off-line system are due to the fact that the "transition period" to
the new technologies is still not over: we cannot count yet on stable
software and hardware environments; there is a lack of new standard tools
and packages replacing the good old ones; the examples of complete
HEP systems based on the new techniques are really few; the "experts"
in the use of the new techniques are really few.