Using Delay States to Report Subsystem Interactions

Not only are multiple address spaces involved in processing transactions, but those address spaces may be part of different subsystems. For example, a CICS® TOR may give control to a CICS AOR who in turn may do a query to IMS™ DLI. Workload management can keep track of subsystems that communicate with each other, and provides the information so that a performance monitor can present the subsystem interactions in processing transactions.

The CICS transaction used in the previous example has a begin-to-end phase and an execution phase. The execution phase could be split among several subsystems, and the delays associated with each distinct subsystem are reported separately. The performance monitor should combine all information by service class by subsystem to provide a sysplex view.

Delay data are presented for as many distinct subsystems as participated in processing each service class. The data are available for both the begin-to-end phase and the execution phase.

A performance monitor could provide a timeline with the various pieces and phases of the work represented differently. In Figure 26, the performance monitor sees that the biggest delays encountered are in the Waiting state and the Continued Local state. The performance monitor could show further information about the Continued Local states by presenting the information found in the execution phase on that system.

Figure 26. Combining state information for a service class.

In the example, the performance monitor shows that there are execution delays attributed to Active, Waiting, and Continued states. Notice that of the 35 delay states reported in the original display as being Continued Local, only 33 of them show up in the execution phase. This is one of the shortcomings of sampling.

In Figure 27, there are two subsystems represented in the execution phase. This means that during this interval, both subsystem A and subsystem B performed work on behalf of service class xyz. Again, notice that there is 1 state sample missing due to statistical anomalies. The performance monitor could determine the specific reason for subsystem B Waiting states from the delay states.

Figure 27. Combining state information across subsystems.