The European Organisation for Astronomical Research in the Southern Hemisphere (ESO), headquartered in Garching,
Germany, operates different state-of-the-art observing sites in Chile. To manage observatory operations and observation
transfer, ESO developed an end-to-end Data Flow System, from Phase I proposal preparation to the final archiving of
quality-controlled science, calibration and engineering data. All information pertinent to the data flow is stored in the
central databases at ESO headquarters and replicated to and from the observatory database servers.
In the ESO's data flow model one can distinguish two groups of databases; the front-end databases, which are replicated
from the ESO headquarters to the observing sites, and the back-end databases, where replication is directed from the
observations to the headquarters.
A part of the front-end database contains the Observation Blocks (OBs), which are sequences of operations necessary to
perform an observation, such as instrument setting, target, filter and/or grism ID, exposure time, etc. Observatory
operations rely on fast access to the OB database and quick recovery strategies in case of a database outage.
After several years of operations, those databases have grown considerably. There was a necessity in reviewing the
database architecture to find a solution that support scalability of the operational databases.
We present the newly developed concept of distributing the OBs between two databases, containing operational and
historical information. We present the architectural design in which OBs in operational databases will be archived
periodically at ESO headquarters. This will remedy the scalability problems and keep the size of the operational
databases small. The historical databases will only exist in the headquarters, for archiving purposes.