z/OS MVS Programming: Workload Management Services
Previous topic | Next topic | Contents | Index | Contact z/OS | Library | PDF


Using ENF Signals to Guide Data Collection

z/OS MVS Programming: Workload Management Services
SC34-2663-00

Using ENF Signals to Guide Data Collection

Because all performance data is continually collected until a significant change takes place in workload management, the performance monitor must know when the collection is reset. Workload management resets its collecting when:

  • A service policy is activated (VARY WLM,POLICY=)
  • A system error occurs.

ENF signals are issued for these events, and for each ENF, a listener is notified synchronously at the start of a change, and at the completion of a change. There is an ENF code qualifier indicating a change started, an ENF code qualifier indicating a change completed, and an ENF code qualifier indicating a change has failed.

When a change is started, workload management captures a last copy of workload data. The performance monitor can listen for the “start” event code, and then invoke IWMRCOLL to get the last available data. The workload data remains the same until either the change completes, or the change fails. Based on the “start” event code, the performance monitor should save this data, end its current interval, and wait for the next code.

Note:
Because the ENF signal is issued synchronously, the listener should complete processing as quickly as possible. For example, if the ENF is issued broadcasting the start of policy activation, the policy activation is held up until all listeners for that ENF have relinquished control.

If the change is completed, the performance monitor should invoke the proper services, as shown in Table 18 . A performance monitor should always complete its last interval with the data collected when it received the “start” event code. If the change failed, the performance monitor can continue to use the data it saved when it received the start event code. Regardless of whether the event has completed or failed, the performance monitor should reinitialize all of its workload activity information. Then the performance monitor should issue:

  • IWMPQRY for a copy of the active policy
  • IWMRCOLL for workload activity information.

This way, the performance monitor does not need to do two different things for completions and failures.

There is also an asynchronous ENF signal issued when the IWMWRQAA information is available. That way, a performance monitor can synchronize its sampling interval for address space states with workload management's sampling interval.

Table 18 shows the ENF event code and its qualifiers. The table also outlines what a performance monitor could do when the ENF code is heard.

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014