Using SMF collected data, it is possible to identify specific intervals when the problem program use of system resources is at an extremely high or low level. By studying the trends in this SMF data, and the relationships among the trends, operations management can establish and enforce its job scheduling procedures. The following examples describe a few potentially useful SMF data-trend analyses for scheduling jobs.
The SMF job initiation and termination records contain the start and stop times of each batch job, job step, TSO/E session, APPC/MVS transaction program, and started task. Using these times, an installation can determine the jobs that are running during the same interval. From a scheduling point of view, a low number of concurrent jobs might indicate the need for establishing more job classes or using more initiators.
The SMF step termination records have the following three time stamps: step initiation time, device allocation start time, and problem program start time. By calculating the differences in these three times, an installation can identify any abnormally long job step initiation.
In addition, an installation can use the SMF output writer and job purge records to track job wait times in both the SYSIN and SYSOUT queues over a given period of time. If the resulting pattern of wait times shows any significant variances, the installation might want to further investigate the problem areas and perhaps alleviate them by rescheduling manpower or changing hardware.
By examining the SMF-recorded job accounting fields (such as department number, project number, and user ID), and the SMF-recorded job initiation time and date fields, an installation can create a fairly accurate picture of its job throughput and turnaround time. For instance, one installation might analyze its throughput by calculating the total number of jobs initiated within each 15-minute interval and categorizing its jobs into test and production jobs.
Such an installation could use SMF to determine the time of day when the largest number of production jobs were going through the system. Then, by limiting the number of test jobs during that time, the installation might improve its production turnaround time.
By summarizing this type of SMF information for all jobs and job steps over a given period of time, an installation can establish its workload characteristics and set specific standards for each job class and priority. An installation can use this information to determine whether it is meeting its service goals.