Learn the common steps involved with identifying a performance problem.
When trying to identify a performance problem, it is important to assess whether the hardware configuration is adequate to support the workload. Is there enough CPU capacity? Is the main storage sufficient for the different types of applications? Answering these questions first, perhaps through capacity modeling techniques, prevents needless effort later.
With an understanding of the symptoms of the problem and the objectives to be met, the analyst can formulate a hypothesis that may explain the cause of the problem. The analyst can use commands and tools available with i5/OS™ and the Performance Tools licensed program to measure the system performance.
Reviewing the measured data helps to further define the problem and helps to validate or reject the hypothesis. Once the apparent cause or causes have been isolated, a solution can be proposed. When you handle one solution at a time, you can redesign and test programs. Again, the analyst's tools can, in many cases, measure the effectiveness of the solution and look for possible side effects.
To achieve optimum performance, you must recognize the interrelationship among the critical system resources and attempt to balance these resources, namely CPU, disk, main storage, and for communications, remote lines. Each of these resources can cause a performance degradation.
Improvements to system performance, whether to interactive throughput, interactive response time, batch throughput, or some combination of these, may take many forms, from simply adjusting activity level or pool size to changing the application code itself. In this instance, an activity level is a characteristic of a subsystem that specifies the maximum number of jobs that can compete at the same time for the processing unit.