Abstract: A method and apparatus for automatic discovery of databases that determines what databases are installed at a client site, interrogates each database to determine what files or tables are present and what storage is associated with each, groups the databases into the smallest atomic units that can be backed up while preserving database integrity, and presents this information to backup procedures. In a preferred embodiment, a discovery process is performed by a script for each client to be backed up. Each script causes the creation of a temporary file describing each database. The temporary files are normalized into a common format which is then passed to a grouper program that analyzes the temporary files and produces a data descriptor file describing the smallest atomic units that can be backed up while still preserving database integrity at that client site. This data descriptor file is presented to a work item generator for scheduling backup procedures.