Database

All operations in PanDA are handled and checkpointed through a central database that is shared by all the PanDA and JEDI servers. ATLAS operation at scale (O(1M) jobs per day) is performed using an Oracle backend and several optimizations have been build over the years. There is also a more limited experience and support for MySQL/MariaDB. A few PanDA instances are operated by different experiments, but the scale is several orders of magnitude lower requiring far less optimization. Also not the full functionalities of PanDA and JEDI are used. Currently PanDA SQL code is written and tested for the ATLAS Oracle case and is translated to MySQL dialect through a in-house function. I.e. there is no usage of ORM layers as SQL Alchemy.

You can find more information in the following sections.

• Client configuration
• Database server requirements
• Entity relation diagrams
• Partitioning
• Archival strategy
• Exporting Oracle Database Schemas to PostgreSQL
• Database Administration