What is a performance problem?

There are many views on what constitutes a performance problem. Most of them revolve around unacceptably slow response times or high resource usage, which we can collectively refer to as "pain." The need for performance investigation and analysis is indicated by, for example:

Slow or erratic response time:
- Service level objectives being exceeded
- Users complaining about slow response
- Unexpected changes in response times or resource utilizations
Other indicators showing stress;
- Monitor III Workflow/Exceptions
- System resource indicators (for example, paging rates, DASD response)
- Expected throughput on the system not being attained

Finally you need to decide whether a given situation is a problem worth pursuing or not. This decision is based on your own experience, knowledge of your system, and sometimes politics. We simply assume for the following discussions that you are trying to relieve some sort of numerically quantifiable performance problems in your system.

Generally, a performance problem is the result of some workload not getting the resources it needs to complete in time. Or the resource is obtained but is not fast enough to provide the desired response time.

The most frequent cause of performance problems is having several address spaces compete for the same resource. These could be a hardware resource (processor, device, storage) or serially usable software resource (catalog, VTOC). While one address space is using the resource, the other address spaces are delayed. Therefore, one key aspect of Monitor III is to make visible who is using what, and who is delayed.