Blocking is the process of grouping records into blocks before they are written on a volume. A block consists of one or more logical records. Each block is written between consecutive interblock gaps. Blocking conserves storage space on a volume by reducing the number of interblock gaps in the data set, and increases processing efficiency by reducing the number of I/O operations required to process the data set.

If you do not specify a block size, the system generally determines a block size that is optimum for the device to which your data set is allocated. See System-Determined Block Size.

You select your record format in the data control block (DCB) using the options in the DCB macro, the DD statement, dynamic allocation, automatic class selection routines or the data set label. Before executing your program, you must supply the operating system with the record format (RECFM) and device-dependent information in data class, a DCB macro, a DD statement, or a data set label. A complete description of the DD statement keywords and a glossary of DCB subparameters is contained in z/OS MVS JCL Reference.

All record formats except U can be blocked. Variable-length records can be spanned (RECFM=DS or VS). Spanned records can span more than one block. Fixed-length records (RECFM=F or FB) can be specified as standard (RECFM=FS or FBS). Standard format means there are no short blocks or unfilled tracks within the data set, except for the last block or track.