When running on rcas, we typically use a legacy csh script named "alljobs". It parses the job configuration file named "catalog" and dispatches a single job on the target node, which can be an interactive node if run interactively, or a batch node if submitted to a queue. The alljobs script expects to see the following directory structure: a writeable directory with the name of the dataset being produced, and directly under it, a writeable "log" directory, in which it would deposit the so-called token files, which serve two purposes:

• help in sequential numbering of the output files
• in case of multiple input files (such as Hijing event files) allow to map N files to M jobs, thus managing the workload

The catalog file is effectively a table, in white-space separated format. Each line begins with the dataset name which is a three-letter acronym of the site name (and thus either rcf or pds) followed by a 4-digit serial number of the set. The alljobs script expects to find a directory named identically to the dataset, under the "job directory", which in the current version of the script is hardcoded as /star/simu/simu/gstardata. This, of course, can be improved or changed.

The last field in each entry is used to construct the so-called tag, which plays an important role: it effectively defined the location of the Monte Carlo data in the HPSS, when the data is sunk there (this is done by a separate script). In addition, it also defines keys for the entries in the FileCatalog (reconstructed data). The alljobs script creates a file of zero length, with a name which is a period-separated catenation of the word "tag" and the contents of the last column in the line.

Here are the contents of the catalog file as it existed from the late 1990-s to the end of 2006

Here is the actual version of the file used in the 2007 runs:

When submitting jobs on the Grid, most of the functionality in alljobs is redundant. The simplified scripts can be found in the "Grid-Friendly" section of this site.