Global tables

The JSOC catalog will maintain a set of global tables that define the structure of data records belonging to each series, and contain information shared by all record of the series, such as tape storage group, default online retention time and whether records from the series should be archived to tape.

An example of what these tables might look like is shown in Figure 2.

Figure 2: Example illustrating the structure of the global tables in the JSOC catalog.

• The master_series_table contains a list of all JSOC data series with descriptions of the series, information about when and by who they were created, in addition to information about the storage policy for the series.

• The master_keyword_table contains a list of all keywords for all JSOC data series. Each row in the table describes things like the name of the keyword, the data type, default output format and physical unit of its value and whether the keyword is "inherited" by following a link to another data record or whether it is stored as a simple value in the "series table", which is the main table holding actual the keyword values for all records belonging to it (see below). Additional information might include whether the database should maintain an index on this keyword value. This is done to speed up database queries with conditions involving this keyword value.

• The master_link_table contains a list of all links for all JSOC data series. Each link has a name and a target series to which they point, and can be either static, i.e. pointing to a specific data record in the target series defined by (record number, version), or it can be a "query link". A "query-link" is can be represented as (link name, series name, query) tuple. The intention of the later form is that the link can be bound/resolved dynamically by evaluating the query condition either at creation time or when the data record is opened for reading. This can for example be used to create a link that automatically resolves to the most recent data record of a data series, e.g. the most recent version of a calibration table. The syntax used in query links is TBD.

• The master_data_table contains a list of the data segments for records belonging to JSOC data series. A data segment is described by its name, type, physical unit, dimensions and the storage protocol used when accessing the contents of the data segment in secondary storage.

The master tables above merely describe the structure of data records belonging to data series and are therefore reasonably small. The number of rows in these tables is proportional to the number of data series [which is probably no more than a few hundreds, maybe a thousand.]. The bulk of the JSOC catalog is a set of tables, one for each series, containing the actual keyword values for all data records. In addition, a special sequence table is maintained for each series. A sequence is a special database table type containing a counter that can be read and incremented in an atomic operation. It is used to guarantee that unique record numbers are generated even when several modules concurrently are creating new records belonging to the same series.