CICS performance analysis techniques

A number of techniques are available for analyzing CICS® performance.

There are four main uses for performance analysis:
  • You currently have no performance problems, but you want to adjust the system to give better performance.
  • You want to characterize and calibrate individual stand-alone transactions as part of the documentation of those transactions, and for comparison with some future time when, perhaps, they start behaving differently.
  • A system is departing from previously identified objectives, and you want to find out precisely where and why. Although an online system might operate efficiently when it is installed, the characteristics of the system usage can change and the system might not run so efficiently. This inefficiency can usually be corrected by adjusting various controls. Some adjustments usually need to be made to any new system when it goes live.
  • A system might or might not have performance objectives, but it appears to be suffering severe performance problems.
If the current performance does not meet your needs, consider tuning the system. To tune your system, you must perform the following tasks:
  1. Identify the major constraints in the system.
  2. Understand what changes could reduce the constraints, possibly at the expense of other resources. Tuning is usually a trade-off of one resource for another.
  3. Decide which resources could be used more heavily.
  4. Adjust the parameters to relieve the constrained resources.
  5. Review the performance of the resulting system in the light of these criteria:
    • Your existing performance objectives
    • Progress so far
    • Tuning effort so far
  6. Stop at this point if performance is acceptable; otherwise do one of the following actions:
    • Continue tuning
    • Add suitable hardware capacity
    • Reduce your system performance objectives.
The tuning tasks can be expressed in flowchart form as follows:
Figure 1. Flowchart to show rules for tuning performance
First, understand the performance objectives. Second, monitor the system following a measurement and evaluation plan, to establish objectives, resource contention, and predictions. Third, check whether the performance objectives are currently being met. If the performance objectives are currently being met, continue to monitor the system following your plan. If the performance objectives are not currently being met, identify a major resolvable resource contention, then devise a tuning strategy that minimizes usage of resource or expands the capacity of the system, then identify the variables, then predict the effects, then make the change. Recheck whether the performance objectives are currently being met, and repeat these steps again as necessary, until they are being met.


dfht32q.html | Timestamp icon Last updated: Monday, 5 June 2017