Experience with the Open Source based implementation for ATLAS Conditions Data Management System

Conditions Data in high energy physics experiments is frequently seen as every data needed for reconstruction besides the event data itself. This includes all sorts of slowly evolving data like detector alignment, calibration and robustness, and data from detector control system. Also, every Conditions Data Object is associated with a time interval of validity and a version. Besides that, quite often is useful to tag collections of Conditions Data Objects altogether. These issues have already been investigated and a data model has been proposed and used for different implementations based in commercial DBMSs, both at CERN and for the BaBar experiment. The special case of the ATLAS complex trigger that requires online access to calibration and alignment data poses new challenges that have to be met using a flexible and customizable solution more in the line of Open Source components. Motivated by the ATLAS challenges we have developed an alternative implementation, based in an Open Source RDBMS. Several issues were investigated land will be described in this paper: -The best way to map the conditions data model into the relational database concept considering what are foreseen as the most frequent queries. -The clustering model best suited to address the scalability problem. -Extensive tests were performed and will be described. The very promising results from these tests are attracting the attention from the HEP community and driving further developments.