What is a Partitioned Dataset(PDS)?

What is a Partitioned Dataset(PDS)?

When you create a sequential dataset(PS), it occupies a large amount of storage space(minimum 50,000 bytes). Sometimes for doing trivial tasks,  storing only a few records, you would not be utilising the entire space available in a sequential file. Thus most of the space in the Sequential file is wasted and is blocked.

IBM provides a way to split or partition up the space in a sequential data-set. IBM Software engineers invented the Partitioned Dataset, often called PDS. Each part is called a Member. A Partitioned Dataset can have many members(parts). Each member behaves like a sequential file on it own.

How does the PDS then keep track of its members? A Partitioned Dataset(PDS) includes a directory(like a telephone directory), which is a diary or journal, where MVS Operating system keeps track of the members in a PDS. The members of a PDS can be scattered haphazardly anywhere in the vast computer storage space. Well-then, you consult the directory to find out the members in the PDS. Just like in a book, the index tells you, the keyword you are looking for, and the page no. where it can be found, the same way, the directory tells you the member name, and the computer memory address where you can find it. Thus, the directory maintains the list of member-names and pointers to the actual physical place in Computer Memory where they are stored.

Not just this, but the directory also keeps track of its statistics, like when each member was created, when was it last changed/viewed by somebody, how much space it occupies, which TSO-userid created it and bla bla..

On the Windows Operating System, we generally keep related files together in a folder. On the same lines, you keep related members together in one PDS. Thus, a Partitioned Dataset is Mainframes-counterpart of Folders on Windows Operating System. Folders contain many related files. A PDS contains many related members.

When create this special type of partitioned sequential file – Partitioned Dataset, you need to decide, how big its directory is going to be. The bigger the directory, the more no. of partitions(members) it will support. You express this in terms of Directory Blocks. Generally, 1 directory block can store information about 5 members. The directory always remains fixed, can’t grow bigger on the fly, so you need to careful when specifying the directory blocks, and you have to do it in advance.


To sum up, a Partitioned Dataset(PDS) is a specialised type of sequential file, which is divided into members, each member being a sequential dataset by itself. Thus, a PDS acts like a library which houses several related files.