An approach to evaluating time sharing systems: MH-TSS a case study

The authors conducted a benchmark measurement of the Murray Hill Time Sharing System (MH-TSS) running on a Honeywell 6000. The object of the test was to duplicate the load normally present on the Murray Hill production system, and measure the system's behavior before and after a major software release and a major hardware improvment. Five different load levels, from 30 to 90 users, were measured for each configuration. This paper discusses the methods used in the design of the experiment and in the analysis and interpretation of the results.Several measurement tools were used in this test. The event trace collection facility of MH-TSS was used for the benchmark measurement and for the design and fine tuning of a scrint representing the normal load at Murray Hill. A commercially available H6000-specific terminal simulator was used to feed these scripts to the system. The batch background system was loaded by a stream of synthetic jobs, matched in resource usage characteristics to a set of jobs chosen at random from the job stream of the production system.The event trace data gathered at various load levels under the three software and hardware configuratios were analyzed using two techniques employing a state transition representation of program behavior and system response. The result was a set of data which documents the expected performance improvements for the new software and hardware being installed at Murray Hill, and which suggests the expected growth potential for MH-TSS.