MONWRITE Data and Upgrades

Last revised: 2022-05-10, BKW

Upgrades happen. Clients move from one IBM Z or LinuxONE CPC to another, or from one z/VM level to another, or from one guest operating system level to another, or from one middleware version to another, and the like. This is all a part of life.

IBM wants clients' upgrades to be successful. This means IBM wants clients to be well prepared for upgrades. As a performance specialist, I want clients to be well prepared for the performance aspects of their upgrades. Hence, this tips article.

For IBM to help with performance situations in upgrades, IBM needs performance data from both before and after the upgrade. Proceed like this:

  1. Identify one or more processing periods that are important to your business. Maybe it's week-end, or month-end, or the overnight reprocessing window. You know your workload best.

  2. For each such processing period, do this:
    1. Collect only one hour of MONWRITE data,
      1. With one-minute sample intervals,
      2. With high-frequency state sampler interval one second,
      3. With all sample domains enabled,
      4. With all event domains enabled except seeks and scheduler,
      5. With CPU MF counters included in the collection.

      For help, visit this MONWRITE article and this CPU MF collection article.

    2. Concurrently with collecting the MONWRITE data, if you are running guest operating systems that emit their own performance data, collect that too. For instructions about collecting such data, refer to manuals or specialists associated with those operating systems.
    3. Concurrently with collecting the MONWRITE data, collect something that describes your workload's own view of how successful it is being in doing its work. This probably includes some measure of rate of work, such as the rate at which the workload is processing credit card charges. This probably also includes how much elapsed time is used in processing a single element of work, again, such as a credit card transaction. IBM will be grateful for any data that relates the workload's own performance experience to the resource consumption described in MONWRITE data.

Handy tip: The above data is certainly useful in the context of upgrades. But it's also useful for far more than just that. If you are not routinely collecting and archiving specimens of performance data, now is a great time to start. One never knows when such data will come in very handy.

Thanks, and we look forward to your successful upgrades.