UNPARKING MEDIUM Experiment

In November 2017 IBM shipped the PTF for APAR VM66063. This new function APAR introduced CP SET SRM UNPARKING and offered three unparking models: LARGE, MEDIUM, and SMALL. The descriptions of the models are available here.

Clients who installed the PTF and did nothing further would be running with UNPARKING LARGE, which is how z/VM behaved prior to applying the PTF. To switch to a different model, the client would exploit either the new z/VM system configuration file statement or the new CP command, selecting either UNPARKING MEDIUM or UNPARKING SMALL.

If you have tried UNPARKING MEDIUM, I am interested in your experience. What led you to try it? How did you measure the behavior change? What factors in your collected measurement data did you consider in deciding either to stay with UNPARKING MEDIUM or to return to UNPARKING LARGE?

If you haven't yet tried UNPARKING MEDIUM, maybe you'd like to give it a try and share your experience. Here's how to proceed:

  1. Pick a time when workload on your CPC is suitable. Maybe you would like to pick a time when workload on some LPARs is light and on other LPARs it's busy. You decide.
  2. Make sure all LPARs on your CPC are running with Global Performance Data Control turned on. You can do this on the SE's "Change LPAR Security" screen. Turn on all the GPD checkmarks and then click "Change Running System".
  3. Start collecting MONWRITE data. Be sure to issue CP MONITOR EVENT ENABLE PROCESSOR, so we can see the D5 R16 MRPRCPUP records. Also be sure to run with CPU MF host counters enabled; if you need help with that, visit our CPU MF counters tips page.
  4. Start collecting whatever data it is you use to assess application behavior, such as transaction rate, transaction response time, or whatever it is you use.
  5. Collect for a few sample intervals.
  6. Switch all your LPARs to medium unparking. On every LPAR, issue this command: CP SET SRM UNPARKING MEDIUM.
  7. Collect for a few more sample intervals.
  8. Switch all your LPARs back to large unparking. On every LPAR, issue this command: CP SET SRM UNPARKING LARGE.
  9. Collect for a few more sample intervals.
  10. Stop collecting your application performance data.
  11. Stop collecting your MONWRITE data.
  12. If you changed any LPARs' Global Performance Data Control checkmarks, change them back.

Sending Your Data to IBM

Because of recent changes to privacy laws, IBM can no longer accept MONWRITE data transmitted as e-mail attachments, sent on USB sticks, or the like. Rather, all data must come to IBM submitted through a PMR and uploaded to ECuRep in accordance with the instructions for sending PMR data.

To send us your data, then, you will have to open a PMR with z/VM Level 2. When you do so, be sure to tell Level 2 that you are sending data for the Medium Unparking study and that the data is for me.

When you send us your data, feel free also to tell us any specific observations or concerns you had during the experience.

For more information about sending data to IBM, see here.