DRA statistics
DRA statistics are contained in the returned PAPL as a result of a DRA TERM request, or in the Control exit routine's PAPL when it is called for DRA termination. This routine is called when the DRA fails or when a previous Control exit routine invocation resulted in return code 4.
The DRA statistics in the returned PAPL are:
- The number of times the MAXTHRD value was reached.
- The number of times the MINTHRD value was reached (only includes the times the value is reached when the thread TCB number is decreasing.)
- The largest number of thread TCBs ever reached during this DRA session. (This is the number of TCBs, not the number of DRA threads, so it is at least the minimum thread value.)
- The time (in seconds) during which the DRA thread TCB count was at the MAXTHRD value.
You can find the field names for the previous statistics in the PAPL extensions for the TERM PAPL and control exit routine PAPL.
Before attempting to evaluate the statistics DRA performance, remember:
- If the DRA is using the maximum number of threads (MAXTHRD), when the DRA receives any new SCHED requests it will make these requests wait until a thread is available.
- As active threads become available (for example, as a result of TERMTHRD call), some of the available threads might be collapsed.
These factors can adversely affect performance, but both improve IMS DB resource availability because fewer DRA threads require fewer IMS DB resources. The IMS DB resources (PSTs) are then available for other BMPs or other CCTLs to use.
Statistics 1, 2, and 4 can serve as a measurement of the two factors, and will help you decide how to balance performance and resource usage. For the sake of this discussion, these statistics are presented solely from a performance point of view (for example, assume only 1 CCTL connected to a IMS DB).
Evaluating the DRA statistics
If statistics 1 and 4 are high, a SCHED request had to wait for an available thread many times. To improve performance, raise the MAXTHRD value.
The impact of statistic 2 on performance can only be estimated if thread activity history is known (the DRA does not provide this history but the CCTL can). If activity is steady, little thread collapsing occurs and statistic 2 is meaningless. If activity fluctuates a lot, statistic 2 can be useful.
- If statistic 2 is 0, thread collapsing might have occurred, but the MINTHRD value was never reached.
- If statistic 2 is not zero, the MINTHRD value was reached and thread collapsing was stopped at those points, thus enhancing performance. Therefore, if you have highly fluctuating thread activity, you can improve performance by raising MINTHRD until statistic 2 has a nonzero value.
Finally, statistic 3 can be useful for adjusting your MAXTHRD value.