STAR Computing GRID Scheduler Development
|
|
STAR Computing GRID Scheduler Development File Catalog |
One of the key feature of the scheduler is the ability to connect to the meta-data/file catalog. Most of the time a user will specify his input in terms of metadata attributes. The architecture of the scheduler allows for different catalogs to be integrated.
More detailed information on how we reached the current interface can be found here. Here we provide a summary.
In the request, a user will write something like:
<input URL="catalog:star.bnl.gov?production=P03ia,trgsetupname=dAuMinbias,filetype=daq_reco_MuDst,magscale=ReversedFullField" />
The part after '?' in the URL is file catalog specific, and will be passed the file catalog module to determine the files to which the query correspond to.
The FileCatalog class represent a file catalog/metadata catalog. The model is the following: the FileCatalog class is given a string, representing the query, and the FileCatalog returns an Iterator to a Collection of PhysicalFiles. The StarFileCatalog is the implementation used for the Star file catalog.
The <input> tag provides many attributes that tell the scheduler how to solve the query. For example, it tells how many files to retrieve, and whether we want all the copies, or
Gabriele Carcassi - page was last modified
 |
 |
 |