Performance modeling and empirical measurements in a system designed for batch and time-sharing users

Abstract

If any design goal is common to all computer system organization schemes, it is that of providing "effective service" both externally to the user of the computational facility and internally with respect to utilization of system resources. Thus, generally speaking, there are at least two dimensions to this design objective. On the one hand, effective service is the external satisfaction of a broad spectrum of user demands. For example, the ideal system might be visualized as one which economically provides a large number of programming languages; machine compatibility with other computers of widely diverse hardware; and rapid computation. On the other hand, effective service is the internal utilization of all system components so as to increase computational efficiency. In this respect, system structures are implemented which strive to maximize sub-system simultaneity and system throughput. For example, a degree of macro-parallelism is attained in many present day systems by allowing a central processing unit (CPU) and input/output controller to share the use of a main memory register, thereby enabling processing and input/output (I/O) to proceed concurrently (for one or several independent programs, depending upon the system software).