The files you create with the z/OS NFS server are contained in MVS data sets. These MVS data sets are record-oriented and can be sequential, direct, VSAM, partitioned, and so forth. These MVS data sets are variable or fixed in record length. UNIX files, however, are byte-oriented and typically written or read at certain offsets in the file.

You can map non-MVS files to most types of MVS data set organizations. However, how the time stamps and file size value are handled depends on the type of MVS data set used, and the file size processing can affect performance. See File size value for MVS data sets.

Direct reads with record format recfm(fbs) or recfm(f) can be fast. In some cases, the z/OS NFS server can determine the physical block addresses from the record offsets. The MVS sequential file organization with recfm(fbs) or recfm(f) on DASD allows for efficient updating or reading at any offset in the file. Other supported MVS access methods (for example, VSAM) can be used if required by a given application but, in general, the sequential file organization is the best choice for files that are used mainly by UNIX clients.