Flexible Storage of Astronomical Data in the ALMA Archive

The requirements for the archiving of ALMA observation data are challenging: Not onlyare the expected rates of observation and monitor data extremely high (0.5 TeraByte/day), there is also the need to archive metadata about projects, proposals, observations, scheduling blocks, etc. in a flexible waythat allows for changes in the structure of these data over the years. The ALMA archive is divided conceptuallyinto three parts: (1) The BulkStore for the veryobservation data, (2) the MonitorStore for monitor data collected byall instruments, and (3) the XMLStore for metadata about observation and monitor data. The entities in the three distinct stores are highlyinterrelated. We will give an overview over the architecture of the ALMA archive with a special focus on XML storage. XML (eXtended Markup Language) was chosen not onlyas format for communicating data in the ALMA computing infrastructure, but also for archiving data, since it provides the required flexibilityneeded bythe ALMA archive: XML is designed to represent semistructured data, i.e. data whose structure is irregular, changing over time or even unknown. This makes it the format of choice for software that has to work over manyy ears, when changes in the underlying data structures are unavoidable.