This topic covers handling issues related to performance degradation of the system.
The performance issues are those where the response time for output is longer than expected time for the execution. The performance can be either due to external software such as databases or hardware or due to any TXSeries region setup.
TXSeries can resolve particular instances of poor performance. If you identify symptoms of poor performance in your system, such as slow response times or storage stress conditions, you need to identify the performance constraints that are causing the problems and take the necessary corrective action. The diagnosis of performance problems is simplified if you have a set of baseline data. This is normally a combination of TXSeries statistics and Performance Monitor data, which needs to be collected from your production system under both normal and peak load conditions. If you experience degraded performance even then, you can collect statistics and other performance data to compare with your baseline. This method readily identifies causes such as a change in the transaction mix. For example, more frequent and complex transactions that are linked to business cycles.
Practically every symptom of poor performance can result from a congested system. For example, if the available disk space cannot meet demand, the queue of transactions that needs file access starts to build up. When the number of incomplete transactions in the system increases, the greater is the demand for paging and swapping space and physical memory. The task dispatcher also uses more processor power scanning the task chains. In addition, the limit that is defined in the Maximum application servers attribute for the region can be increased, and the resulting retries add to the processor overhead.
The result, where the system shows heavy use of all its resources, is typical of system stress. It does not mean that all resources have a problem, but it does mean that a constraint exists that you must find. To do this, you must find out what is affecting task execution.