An experimental comparison of time sharing and batch processing

Abstract

The effectiveness for program development of the MIT Compatible Time-Sharing System (CTSS) was compared with that of the IBM IBSYS batch-processing system by means of a statistically designed experiment. An identical set of four programming problems was assigned to each of a group of four programming subjects. Influences external to the systems, such as the sequence of problem solution, and programmer and problem characteristics, were specified as design factors in the experiment. Data was obtained for six variables (e.g., programmer time, computer time, elapsed time, etc.) which were considered to be definitive of "system effectiveness," and analysis of variance techniques were employed to estimate system differences in these variables after differences due to the design factors had been eliminated. Statistical analysis of the experimental results provided strong evidence of important system differences, as well as a critque of the experimental design itself with implications for further experimentation.