Aren’t Files and Data Sets the same thing?

For many years I used the term file and data set interchangeably when referring to a z/OS data set. However, a few years ago when I was working on a SHARE presentation, I ran across the following from the OS-360 Concepts and Facilities (which was published in 1964):

This changed my perspective and made me realize that there is a difference between a file and a data set; the term data set and its differentiation from a file was intentional from the inception of what we now call z/OS!

Initially z/OS and its precursors (OS-360, MVS etc.) only supported data sets, however with the addition of Unix System Services, z/OS now supports Unix files which are housed under the covers in z/OS VSAM Linear Data Sets.

Therefore, today calling a data set a file may lead to confusion as to whether you mean a USS file on z/OS or an actual z/OS data set.

What about Objects?

We’ve had objects on z/OS since the early 1990s and they are created, accessed, and managed by DFSMS Object Access Method and like USS files they may be contained within data sets. OAM objects can also be written directly to tape and are used in tape virtualization as well.  

However, today when we talk about objects on z/OS we may also be talking about objects in an on-premises or off-premises cloud where the storage server is connected to z/OS through a network connection. With the inclusion of DFSMS Cloud Data Access method in z/OS 2.5 and 3.1 the cloud objects have become directly accessible from native z/OS and are anticipated to play an even greater role in the future of z/OS.

What’s the Difference?

Let’s take a quick look at the characteristics that distinguish sequential data sets from files or objects.

1)     A data set has logical records that have a predictable defined length by using the Logical Record Length (LRECL) parameter and this piece of meta data is saved in the Volume Table of Contents VTOC.

a.      Fixed length records the LRECL is the exact length of each record in the data set.

b.     A Variable length record data set is defined with a maximum LRECL, and individual records may be equal or less than this maximum. Individual records are preceded by a 4-byte Record Descriptor Word (RDW) that contains the length of that record.

2)     A file has no external meta data that describes the length of each record and generally they call their records lines. These lines contain new line characters at the end of each line to delineate the end of the individual line. I can see how this would have made the designers of OS-360 uncomfortable as an invalid insert of a new line character could have caused unpredictable results in early business computing.

3)     The object is a continuous stream of bytes with meta data that describes the length of the entire object. By default, it has no externally organized meta data to describe individual records within the object, however it does have user defined meta data saved external to the object that can describe the contents of the object.

In summary: Data sets have exact meta data saved in the z/OS catalog, VTOC and VVDS and are not files. Files on z/OS are Unix System Services files that are contained in z/OS data sets to allow Unix applications to run on z/OS. Objects can now be accessed directly using DFSMS CDA, are stored in on-prem or off-prem cloud object storage and can have user defined meta data associated with them.  

Today as I work with DFSMScdm and DFSMScda the difference between a z/OS data set, a z/OS USS file and an object in an on or off-premises cloud is always important!

Get familiar with files, sequential data sets, and objects and get started with DFSMScdm and DFSMScda! Find more information about these solutions in the Links provided!

z/OS DFSMS CDA: https://www.ibm.com/support/z-content-solutions/cda/

z/OS DFSMS CDM: https://www.ibm.com/products/zos-dfsms-cloud-data-manager