Sidebar: IBM Datasets

A sequential dataset is like a UNIX flat file, that is, records are read in order, sequentially one after the other. A partitioned file contains several parts and each part has a name. A partitioned dataset is the closest thing to the UNIX directory hierarchy that the mainframe has to offer; however, unlike UNIX's multiple directory tier structure, a partitioned dataset goes down only one level, to its member. When processing a partitioned dataset, each part (or member) is accessed by its name and then read sequentially.

Indexed sequential datasets consist of an index and collection of data records. Records are accessed via a key. Indexed sequential datasets are essentially structured database tables. Direct access datasets are accessed via the actual disk address in which the record exists.