Partitioned data set
A partitioned data set (PDS), colloquially often called library or as a PO file called (partitioned organized) is a file system on IBM - mainframes , which (called members) into individual partitions is divided. Each PDS consists of a directory and the individual PDS members and comprises at least one track.
When creating a PDS, the number of directory blocks is specified and thus the maximum size of the directory is specified.
The naming is the same as for other datasets . The member can be accessed directly by appending the member name in brackets.
Example:TSO.MVS.CLIST(TEST)
The last qualifier in front of the brackets usually indicates the type of member it contains. However, these are pure naming conventions. Often used are e.g. B .:
- CLIST - contains executable TSO commands
- CNTL - contains executable batch jobs ( JCL )
- LOAD - contains executable programs ( load module ) LOADLIB is also often used. The changes LPALIB and LINKLIB represent special cases of the LOAD libraries for system purposes.
- TEXT - contains text files
- COBOL - COBOL -Quelltexte
- PLI - PL / I source texts
- C - C source texts
- ASM - assembler source code
- PARMLIB - configuration files for the operating system and software products
When a member is updated, it is rewritten at the end of the PDS; the space that the member previously occupied cannot be used for the time being. Only a so-called "Compress" (IEBCOPY utility) pushes the members together again and releases the space.
When the directory is full, no further members can be stored in the PDS. This situation is usually resolved with copy and rename. But there is also a freeware program for directory expansion.
With the introduction of System Managed Storage (SMS), a more modern form of management has existed since MVS / XA, the Partitioned Data Set Extended (PDSE). Here the data is managed in 4 kB blocks, and the space freed up when deleting members can be used immediately without compress . The directory size is completely variable. The adoption of PDSE has been very slow due to its integrity issues until recently. With a special job , existing partitioned datasets can be transferred to this new management form.