Scenario: Improving system performance after an upgrade or migration

In this scenario, you have just upgraded or migrated your system and it now appears to be running slower than before. This scenario will help you identify and fix your performance problem.

Situation

You recently upgraded your system to the newest release. After completing the upgrade and resuming normal operations, your system performance has decreased significantly. You would like to identify the cause of the performance problem and restore your system to normal performance levels.

Details

Several problems may result in decreased performance after upgrading the operating system. You can use the performance management tools included in IBM® i and IBM Performance Tools for i licensed program (5770-PT1) to get more information about the performance problem and to narrow down suspected problems to a likely cause.

  1. Check CPU utilization. Occasionally, a job will be unable to access some of its required resources after an upgrade. This may result in a single job consuming an unacceptable amount of the CPU resources.
    • Use WRKSYSACT, WRKSYSSTS, WRKACTJOB, or IBM Navigator for i monitors to find the total CPU utilization.
    • If CPU utilization is high, for example, greater than 90%, check the amount of CPU utilized by active jobs. If a single job is consuming more than 30% of the CPU resources, it may be missing file calls or objects. You can then refer to the vendor, for vendor-supplied programs, or the job's owner or programmer for additional support.
  2. Start a performance trace with the STRPFRTRC command, and then use the system and component reports to identify and correct the following possible problems:
    • If the page fault rate for the machine pool is higher than 10 faults/second, give the machine pool more memory until the fault rate falls below this level.
    • If the disk utilization is greater than 40%, look at the waiting and service time. If these values are acceptable, you may need to reduce workload to manage priorities.
    • If the page faults in the user pool are unacceptably high, you might want to automatically tune performance.
  3. Run the job summary report and refer to the Seize lock conflict report. If the number of seize or lock conflicts is high, ensure that the access path size is set to 1TB. If the seize or lock conflicts are on a user profile, and if the referenced user profile owns many objects, reduce the number of objects owned by that profile.