Performance objectives and priorities depend on user’s expectations. From the point of view of CICS®, these objectives state response times to be seen by the terminal user, and the total throughput per day, hour, or minute.
The first step in defining performance objectives is to specify what is required of the system. In doing this, you must consider the available hardware and software resources so that reasonable performance objectives can be agreed. Alternatively you should ascertain what additional resource is necessary to attain users’ expectations, and what that resource would cost. This cost might be important in negotiations with users to reach an acceptable compromise between response time and required resource.
An agreement on acceptable performance criteria between the data processing and user groups in an organization is often formalized and called a service level agreement.
Common examples in these agreements are, on a network with remote terminals, that 90% of all response times sampled are under six seconds in the prime shift, or that the average response time does not exceed 12 seconds even during peak periods. (These response times could be substantially lower in a network consisting only of local terminals.)
You should consider whether to define your criteria in terms of the average, the 90th percentile, or even the worst-case response time. Your choice may depend on the audit controls of your installation and the nature of the transactions in question.