Using the Central Component › Data Mining

Data Mining

Data mining allows you to easily and quickly pinpoint high consuming objects for performance tuning. These objects include application programs, system programs, DB2 plans, and subsystems. With data mining, you can search and cross reference stored measurement information by using a data mining scope, which is your user-defined selection criteria for job steps. The most current measurement information of a job step matching the scope is accumulated and made available for cross referencing. For example, based on annual CPU consumption, you can spot the top consuming application programs within your scope and then use a simple cross reference command to list all job steps that are using those applications.

The results of two important calculations are presented for all tuning objects:

• Annual Frequency - Statistical information about job steps is collected from SMF and the Server. For all tuning objects, Performance Management Assistant uses this historical statistical information to calculate the Annual Frequency. For example, assume Data Mining has determined that a certain application program is used in 12 different job steps. The Annual Frequency in the application program is computed as the average number of annual executions of these 12 job steps. If a job step was not observed for a whole year, an approximate value is calculated. In any case, the Annual Frequency does not represent the number of calls (for example, to an application program). Instead, the Annual Frequency is based on the number of related job step executions.
• Annual CPU Consumption - For all tuning objects, Performance Management Assistant calculates the annual CPU consumption. For example, assume Data Mining finds a specific application program in 12 different measurements, that is, job steps. The CPU consumption from each of the 12 measurements is multiplied by the annual frequency of the related job step. The 12 individual calculations are added together to provide the Annual CPU consumption of the specific application program.

The calculation of the Annual CPU consumption is based on the measurements. Therefore, the results are influenced by:

• The number of measurements available in Performance Management Assistant and how much of the TOP Scope is covered by the measurements
• The age of the measurements
• The Data Mining Scope definitions
• Changes in hardware or software releases
• Heterogeneous MVS systems in a sysplex

Performance Management Assistant should process for three months prior to using the data mining feature. This period allows time for Performance Management Assistant to gather statistical information from SMF and the Server, which is used for execution frequency calculations and summations.