Measurement and Improvement of Memory Allocation in a Process Computer

Abstract

The overall performance of a process computer system depends on how well the computer resources are dynamically allocated to the various control tasks. It is often possible to greatly improve the performance once the resource allocation is known. This paper presents a software approach for obtaining this information. It concentrates on measuring the two most important resources: the hardware processor and the memory storage. As an example of the benefits that can be obtained, the measurement of a GEPAC 4000 computer used for nuclear power plant control will be discussed. However the techniques and ideas can be used to measure any process computer system. For the system measured, it is shown that there exists unnecessary movement of programs to and from secondary memory. The cause and correction of the problem is given. Measurements of the modified system show that a factor of two improvement in the through-put was obtained. A corresponding drop in secondary storage transfers is shown.