Performing basic system tuning

To tune your system's performance, you need to set up your initial tuning values, observe the system performance, review the values, and determine what to tune.

To begin tuning performance, you must first set initial tuning values by determining your initial machine and user pool sizes. Then, you can begin to observe the system performance.

Set initial tuning values

Setting initial tuning values includes the steps you take to initially configure the system pool sizes and activity levels to tune your system efficiently. The initial values are based on estimates; therefore, the estimates may require further tuning while the system is active. The following steps set the initial tuning values:

  • Determine initial machine pool size

    Tune the machine pool to under 10 faults/second.

  • Determine initial user pool sizes

    Tune user pools so that the sum of faults for all user pools is less than the number of processors times the processors percent busy. For example, in a system with four processors running at 50 percent busy (4 * 50 = 200), you would set the faults to less than 200 faults/seconds.

Observe system performance

To observe the system performance, you can use the Work with System Status (WRKSYSSTS), Work with Disk Status (WRKDSKSTS), and Work with Active Jobs (WRKACTJOB) commands. With each observation period, you should examine and evaluate the measurements of system performance against your performance goals.

  1. Remove any irregular system activity. Irregular activities that may cause severe performance degradation are, for example, interactive program compilations, communications error recovery procedures (ERP), open query file (OPNQRYF), application errors, and signoff activity.
  2. Use the WRKSYSSTS, WRKDSKSTS, WRKACTJOB and WRKSYSACT CL commands to display performance data.
  3. Allow the system to collect data for a minimum of 5 minutes.
  4. Evaluate the measures of performance against your performance goals. Typical measurements include:
    • Interactive throughput and response time, available from the WRKACTJOB display.
    • Batch throughput. Observe the auxiliary input/output (AuxIO) and CPU percentage (CPU%) values for active batch jobs.
    • Spooled throughput. Observe the auxiliary input/output (AuxIO) and CPU percentage (CPU%) values for active writers.
  5. If you observe performance data that does not meet your expectations, tune your system based on the new data. Be sure to:
    • Measure and compare all key performance measurements.
    • Make and evaluate adjustments one at a time.

Review performance

Once you have set good tuning values, you should periodically review them to ensure your system continues to do well. Ongoing tuning consists of observing aspects of system performance and adjusting to recommended guidelines.

To gather meaningful statistics, you should observe system performance during typical levels of activity. For example, statistics gathered while no jobs are running on the system are of little value in assessing system performance. If performance is not satisfactory in spite of your best efforts, you should evaluate the capabilities of your configuration. To meet your objectives, consider the following:

  • Processor upgrades
  • Additional storage devices and controllers
  • Additional main storage
  • Application modification

By applying one or more of these approaches, you should achieve your objectives. If, after a reasonable effort, you are still unable to meet your objectives, you should determine whether your objectives are realistic for the type of work you are doing.

Determine what to tune

If your system performance has degraded and needs tuning, you need to identify the source of the performance problem and make specific corrections.