Dataset level analysis

Edit online

To enable data set level reporting and analysis, you must have the license for the job and data set module.

When you have the license, IBM Z IntelliMagic Vision processes SMF record type 42, subtypes 5 and 6 in the Reduce step. The output is also generated in the form of CSV files in the Reduce step. Requires specification of the interest group name in the Reduce input commands. Use the COLLECT INTERESTGROUP= command to specify the interest group name. The same interest group name must be specified in the Analyze input commands as well.

COLLECT DATASET=(
JOBDETAIL,
SYNC=0, INTERVAL=60,
INCLUDE(MINIORATE=1, MININTENSITY=50) |
INCLUDE(MINIORATE=1), INCLUDE(MININTENSITY=50),
AGGREGATE=(OFF|DSNTYPE|SERVICECLASS|STORAGEGROUP|SSID),
INCLUDE(MINVOLUMEIORATE=0.5,MINSYSSSCHRATE=0.5),
INCLUDE(MINVOLUMEINTENSITY=1.0,MINSYSSSCHRATE=0.5),
MINSYSSSCHRATE=0.0,
SERVICECLASS=(<name1>,<name2>, ...)
);

The JOBDETAIL parameter is used to request information at the jobname and system detail level, instead of the default aggregate over all jobs and datasets. When this option is enabled, the MINIORATE and MININTENSITY parameters are applied to each job-data set-system combination.

The default sync value is 0. The recommendation is to use the same sync value as used for processing the RMF data. For more information, see the SYNC subcommand in the Reduce COLLECT command.

The INTERVAL parameter specifies the length in minutes of the intervals IBM Z IntelliMagic Vision uses to evaluate data set statistics. The specified interval length must be at least as large as the SMF recording interval.

If the generation of the type 42 records is synchronized, see the Synchronized interval accounting and Nonsynchronized interval accounting in SA22-7630, z/OS MVS System Management Facilities. Do not use this INTERVAL parameter; IBM Z IntelliMagic Vision uses the SMF recording interval.

See SYNC and INTERVAL parameters for more details on specifying interval and sync values.

The AGGREGATE specifies at which level of detail aggregated data set information is maintained. Aggregated data set information is maintained in addition to the individual datasets. The benefit of aggregated data is that it covers the activity from all data set records, not just the activity from data that meets the minimum I/O rate or minimum activity criteria.

By default, if the COLLECT DATASET command is present, but the AGGREGATE parameter is not included, the aggregated output is summarized based on DSS serial and data set type. Other summaries can be requested as follows.

AGGREGATE=(OFF|DSNTYPE|SERVICECLASS|STORAGEGROUP|SSID),
  • AGGREGATE = NO|OFF- Disables the creation of the aggregated data.
  • AGGREGATE = DSNTYPE- Creates the default output, with activity summarized by DSS serial, data set type and data set description. You can see activity for "VSAM RLS", or PDSE, of "KSDS Linear datasets".
  • AGGREGATE = SERVICECLASS - Adds the service class to the aggregation list.
  • AGGREGATE = STORAGEGROUP - Adds the volume group and storage group to the aggregation list, thus giving more details. However, the corresponding file also becomes large.
  • AGGREGATE = SSID - Adds the SSID to the aggregation, so you can track activity down to the LCU/LSS (channel group). This gives the most detail, but the corresponding database file will be large.

With the optional command SERVICECLASS, you can specify a selection of service classes for which type 42 records must be processed. Wildcard * can be used for specifying service classes. If this command is used, the Aggregated data is also limited to these service classes. The selection that is specified here also applies to type 30 records.

When requesting data sets level reporting, you can use two controls to limit the amount of data set detail level information aggregated data is not influenced.

  • MINIORATE specifies the minimum I/O rate that a data set must show to be included in the data set level reporting.
  • MININTENSITY specifies the minimum value for the product of I/O rate and response time. The recommendation is to start with a value between 50 and 100. You can then evaluate if that gives you the data set level visibility that you need. The lowest possible value that you can consider is 1.
  • If there is a specification of both MINIORATE and MININTENSITY, you can combine the commands in 2 ways:
    • INCLUDE (MINIORATE = 1, MININTENSITY = 50) Is interpreted as: I/O rate must be > 1 OR intensity must be > 50
    • INCLUDE (MINIORATE = 1), INCLUDE (MININTENSITY = 50) Is interpreted as:

      I/O rate must be > 1 AND intensity must be > 50

The setting of these parameters can have a substantial impact on how large the file with data set level statistics.
Note: If you specify JOBDETAIL the output files can become either smaller or larger; there will be more entities, but with the MINIORATE and MININTENSITY parameters the resulting number of entries can become less.

MININTENSITY is an expression of how much response time is consumed by a data set and for that reason it may give you better control over data set level reporting than MINIORATE, which does not take the service time into consideration.

To control data set selection based on a data set response time consumption of 50 msec per second or more, or an I/O rate greater than 10, you would code MINIORATE=10, MININTENSITY=50.

Volume level reporting is also influenced by the COLLECT DATASET command. The MINSYSSSCHRATE, MINVOLUMEIORATE and MINVOLUMEINTENSITY parameters define thresholds for volume activity and can have a substantial impact on the file sizes for volume level data, which is stored in the Volume level DFSMS I/O statistics and Job metrics in the high response time array tables.

  • MINSYSSSCHRATE specifies the minimum uncaptured I/O rate that a volume must show to be included in the volume level reporting.
  • MINVOLUMEIORATE specifies the minimum I/O rate that a data set must show to be included in the volume level reporting.
  • MINVOLUMEINTENSITY specifies the minimum value for the product of I/O rate and response time, that a data set must show to be included in volume level reporting.

Because the dataset I/O rate and the uncaptured I/O rate are not positively correlated, both MINSYSSSCHRATE and MINVOLUMEIORATE is required to pass a threshold they can result in suppressing the most interesting data. Therefore, it is highly advisable to combine these thresholds in statements

  • INCLUDE(MINVOLUMEIORATE=0.5, MINSYSSSCHRATE=0.5)

    dataset I/O rate must be > 0.5 OR uncaptured I/O rate must be > 0.5.

A similar consideration applies to the combination of MINSYSSSCHRATE and MINVOLUMEINTENSITY.

When the reduce step is used under z/OS, predefined volume level thresholds are active unless thresholds are explicitly defined in the reduce command. These predefined thresholds are:
INCLUDE(MINVOLUMEIORATE=0.5, MINSYSSSCHRATE=0.5)
INCLUDE(MINVOLUMEINTENSITY=1.0, MINSYSSSCHRATE=0.5)

When it is performed under Windows, by default all data is written.

Detail and aggregated data set information can be viewed in the IBM Z IntelliMagic Vision Web Reporter. Charts are available from Focal Point Datasets and from many drill downs.

The Volume level DFSMS I/O statistics are available in the IBM Z IntelliMagic Vision Web Reporter, in the Volumes section under Jobs and Data Sets.