HFTSINTVL — Specifying the high-frequency throughput statistics interval

The HFTSINTVL parameter specifies the time interval, in seconds, for writing SMF type 98 records, which record high-frequency throughput statistics (HFTS). The value specified is ignored when the WIC parameter is specified. IBM® z/OS® Workload Interaction Correlator uses a 5-second interval, starting at the top of the minute and for each 5-second multiple that follows, for SMF type 98 records.

The supported HFTSINTVL values are 5, 10, 15, 20, 30, and 60 seconds.

When you specify a HFTSINTVL value, SMF type 98 records are written every 5 seconds, starting at the top of the minute and for each 5-second multiple that follows, for one minute at 0, 15, 30, and 45 minutes past each hour. For all other minutes during each hour, SMF type 98 records are written starting at the top of the minute and for each HFTSINTVL multiple.

For instance, HFTSINTVL(20) writes 216 records every hour in the following manner:
  • Minutes 0, 15, 30, and 45: Write 1 record at the top of the minute and for each 5-second multiple that follows, for a total of 48 records.
  • The other 56 minutes: Write 1 record every 20 seconds at the top of the minute and for each 20-second multiple that follows, for a total of 168 records.

IBM recommends a HFTSINTVL value of 5 seconds.

The NOHFTSINTVL parameter disables the HFTS interval and prevents the collection of type SMF type 98 records. This is the default value.

SMF collects type 98 records only when both the HFTSINTVL parameter and TYPE(98) parameter are specified. Specifying NOHFTSINTVL along with TYPE(98) allows you to disable the collection of type 98 records without changing the TYPE setting.

IBM benchmarks did not detect additional measurable CPU overhead due to collecting HFTS data.

IBM recommends configuring Hardware Instrumentation Services (HIS) to collect hardware counters as described in Collecting CPU MF (Counters) on z/OS. Collecting hardware counters enriches the data collected in SMF type 98 subtype 1 records.