Implementing an analytic model for customer information control systems (CICS)

In most installations, it is important that response time consistently meets negotiated service levels for online, interactive systems. As a result, the capacity planning and performance management disciplines have become important functions to these installations. This paper discusses one of the tools used in capacity planning. Initial methods of capacity planning ranged from “seat of the pants” analysis to complex trending techniques. Because it is often necessary to take into account not only the growth pattern of various capacity indicators, but also the interrelationship of these indicators, analytic queuing models have become an accepted method of analysis in the computer measurement arena. The analytic queuing model discussed in this paper is a mathematic representation of the physical parts of a computer system. It is important to note that such an analytic model can only evaluate a finite number of elements or pieces of the system. As a result, when actual measured values are compared with those calculated by the model (the process of model validation) discrepancies can result because of elements that are not explicitly represented in the model. Many of the influences that are not represented by the mathematic model can be removed by tuning the system to remove bottlenecks. When using a model for capacity planning, it is assumed that the data used to develop the model are derived from a system that is “moderately tuned.” A moderately tuned system is one in which the majority of the bottlenecks or other influences not represented by the model have been removed. When the model is used to predict the responsiveness and throughput of the system into the future, it is assumed that any negative influences not represented in the model will be removed in a timely fashion. In this paper, we will review the impact of such tuning issues on attempts to validate the model against actual data. The analytic model discussed in this paper is for IBMs Customer Information Control System (CICS). The model was published in 1982 in the IBM Systems Journal. In the following sections, we will review the model that was presented in the IBM Systems Journal, present extensions required to implement the model for MVS systems, and discuss the aspects of CICS performance that are not represented by analytic models based on traditional SMF and RMF data.