Previous Topic: Interpreting SMF Statistical Records - Step APCXACALNext Topic: Alert Part


Statistical Part

The statistical part reads all statistical measurement records from SMF. Now an attempt to find the job step in the job file is tried. If the record is available, the statistical update follows. If the JCL program name is changed, step APCXACAL detects the change, override the JCL program name, and delete all existing statistical information.

The condition code of the job step is checked. By default, condition codes of 0 or 4 are interpreted as valid for further statistical calculations. If you require additional condition codes to be defined as valid, please contact your first level support.

If the job step record is not already created, the SMF record is analyzed. The condition code of the job is checked and if it is not valid, the newly created job step record only has an increased ccode (condition code) count. All other information from the SMF record is discarded. However, if the condition code for the job step is valid, the job step SMF information is stored on the job step record and the measurement count is set to one.

If the job step record was previously created and the current condition code is not valid, the ccode count is increased and all other SMF information is discarded. If the job step condition code is valid, the runaway test is performed.

The runaway test is a statistical test used to detect runaway situations in the current SMF consumption values for the job step. The implemented test was first described by Nalimov and, simplified, is a check of the SMF measured value against the sum of the average of these values (as maintained by PMA) plus twice the value of the standard deviation. If the measured value is above this sum, a runaway situation with over 95% probability has been detected.

In the runaway test, the SMF value for service units is tested first. If no runaway situation is found, the SMF value for elapsed time of the job step is tested. After the runaway check, the alert thresholds are checked against the SMF consumption values. Then, if no runaway situation is found for the elapsed time value and no threshold alert situation is found, the statistical values maintained by Performance Management Assistant is updated for

Next, the current SMF measurement data is stored in the history table and the measurement counter incremented by one.

If a runaway situation is detected for either service units or elapsed time or an alert threshold situation is detected, the statistical values and current SMF measurement data is updated as described above and the alert part of APCXACAL is performed.