Data set organization (DSORG) specifies the organization of an MVS data set. The z/OS NFS data set organization attribute can be physical sequential (ps), direct access (da), or VSAM. z/OS NFS server supports three types of VSAM files: key-sequenced (KSDS), entry-sequenced (ESDS), and relative record (RRDS). However, keyed access and relative-number access are not supported by the NFS protocol.

The DSORG attribute is ignored for directory-oriented NFS client commands. The data set type (DSNTYPE) specifies whether a PDSE or PDS is to be created when the z/OS NFS receives a mkdir command from the NFS client. A PDSE is created when library is specified, and a PDS is created when PDS is specified. You cannot create another PDS (or PDSE) within a PDS (or PDSE).

Another MVS data set type is extended (EXT), which identifies an extended format data set. With z/OS NFS, these DFSMS-managed data sets are allocated based on specification of a data class (DATACLAS). An appropriate data class must have been defined on the MVS system for an extended format data set to be allocated. One of the values specified when creating an MVS data class is the volume count which determines the number of stripes allocated for an extended format data set. If automatic class selection (ACS) routines have been written to allocate extended format data sets based on such criteria as naming conventions, it would not be necessary to specify a DATACLAS to allocate an extended format data set through the z/OS NFS server.

The hierarchical file system (HFS) and z File System (zFS) are other data set types. They must be DFSMS-managed and be mounted within the z/OS UNIX subsystem. While an HFS or zFS file cannot be allocated using z/OS NFS, they can be accessed through the z/OS NFS server.

Data set organizations and data set types are generally selected based on user and application requirements. However, performance should be considered as one of these requirements. An important factor to consider when evaluating performance is the size of the files to be accessed. For smaller file sizes, we can differentiate between the overhead costs associated with creating or opening a file for output. As the file size increases, this impact becomes less of a factor in the transfer rates.