Archival strategy

PanDA itself does not depend on completed jobs to be archived, but it is relevant for monitoring and audits. ATLAS chooses to keep all of their historical data, but does a yearly archival exercise for some of the tables. ATLAS has two databases and different schemas to implement a two step archival as shown in the diagram. As mentioned before, the archive database is never/rarely accessed and does not need to be very performant. Alternative strategies could be envisioned to avoid having a separate archive database.