Last update:
Apr 1, 2000
|
|
|
|
The HERA-B Database Management for Detector Configuration, Calibration, Alignment, Slow Control and Data Classification
|
Antonio Amorim1,
Vasco Amaral2,
Umberto Marconi3,
Tome Pessegueiro2,
Stefan Steinbeck4,
Antonio Tome5,
Vincenzo Vagnoni3,
Helmut Wolters6
- DESY and LIP/ Univ. Lisboa
- LIP
- INFN Bologna
- Univ. of Hamburg
- LIP/Univ. Beira Interior
- LIP Coimbra / UCP
Speaker:
Antonio Amorim
The measurement of the CP violation angles at HERA-B, together with an acceptable event
storage rate implies complex trigger processing and online reconstruction.
The developed system broadcasts the configuration, calibration and alignment information
simultaneously to large numbers of machines while simultaneously making it available through a tree
of cache database servers. The online calibration and alignment processes store data on active
servers, which propagate update messages notifying the event stream.
The scheme for correlating event data and information in the databases accommodates several
versions of offline reprocessing associated with improved calibration and alignment data. Offline
access to the databases is achieved by a replication mechanism that isolates the online servers
from the load fluctuations induced by offline processing.
To avoid the presence of too many small objects in the slow-control databases, the system allows
storage of changes in individual control data with time. Upon request, the history of sets of
values is re-clustered on the database servers before it is sent to the clients.
The system relies on the Berkeley DB (Sleepycat Soft), and includes a data definition layer (MIZZI)
developed for the experiment while we implemented the associated client/server layers together with
and object interface.
We have set a database schema improving process, in close contact with the sub-detector experts, by
making use of UML design.
With an event rate that prevents the efficient use of event tag databases, and the runs extending
for periods in which conditions can change, we have introduced, in the data quality databases, the
concept of tagged data sets that are managed by a relational database system.
After some prototyping projects, we have defined the performance requirements of the database GUI’s
and are implementing the JAVA and TCL/Tk interfaces, optimizing the CPU intensive data manipulating
methods using the C/C++ native interface.
Presentation:  | Short Paper:  |
|
Bulletins
| Home | Bulletins | Committees | Scientific Program | Docs by topics | Social Event | Conference Location | Secretariat |
Privacy Policy |