SMF record format

Depending on the feedback options you select, RMF™ can write the SMF records to the SMF data set, use the data in the record to generate a printed report, or both. Regardless of the options you select, the format of the SMF record is the same.

Each SMF record that RMF generates consists of the following sections:

  1. SMF common header, which identifies the record length, the record type, the time and date, the SMF system identifier, the subsystem identifier (always RMF), and the record subtype (if required). It also describes the other sections in the record. Each section is identified by its offset, the length of the section, and the number of such sections in the record. These offset/length/number triplet pointers define the structure of the rest of the record.
  2. RMF product section, which includes information such as the RMF version number, the start time of the interval, the length of the interval, the length of the sampling cycle, and interval synchronization data. The RMF product section is the same in all records.
  3. Control section, which contains general one-time data for RMF to use to produce any requested report. The contents of the section depend on the record type. Some records do not require a control section, while others require more than one.
  4. Data section, which includes the specific data gathered during the interval. The format and the number of the data sections depend on the record type and the data collected. For example, there would be one data section for each device included in the type 74 record, I/O device activity.

With this format, the SMF records that RMF generates can change to incorporate any new or modified data without creating incompatibilities. The key factors in allowing for compatible change are the grouping of similar data in one section and the use of the offset/length/number triplet pointers to access the data stored in each section. Figure 1 shows the general format of the SMF records that RMF generates. The figure shows both the pointer structure and the storage layout for the sections.

Also, you can access fields in the SMF common header and the RMF product section by either a general name or a specific name. For example, you can access the interval start time in a type 70 record by either its general name (SMFIST) or its specific name (SMF70IST). Thus, code that processes all records can use the general name while code that processes only a specific record type can use the specific name.

Figure 1. SMF Record Format
This figure shows the SMF Record Format and is explained in the surrounding text.

If your installation has existing data reduction programs that use SMF record input, check the SMF record formats carefully to determine what changes are required. Note that using the SMF record mapping macro instructions supplied by RMF is the most flexible way to access the contents of the SMF records your programs require. When you use the mapping macros, usually only a re-assembly of your program is required to incorporate changes to the record format.

The SMF record mapping macro instruction is ERBSMFR. Its format is: 
ERBSMFR(nn1[,nn....1])

where nn identifies the type(s) of the SMF record(s) you want to map. Note that the parentheses are required only when two or more SMF record types are specified.

If you specify ERBSMF, the macro generates a mapping of the SMF common header and the RMF product section using only the general names.

The mapping macros reside in SYS1.MACLIB.

Because RMF can generate spanned SMF records – particularly when I/O device activity is measured – correct DCB parameters are important. Do not override the DCB parameters in the data set label by specifying DCB parameters on JCL statements. However, when using unlabeled tape the JCL describing an input SMF record data set should specify RECFM=VBS and a logical record length (LRECL) that is at least equal to the length of the longest record.