Computer System Design Using a Hierarchical Approach to Performance Evaluation.
Abstract
The concept of a hierarchy of system models for the performance evaluation of computer systems is introduced. The characteristics and construction of such a hierarchy are discussed. Since it consists of models that span a wide range of complexity and cost, such a hierarchy is a very useful tool in the cost-effective design of computer systems. A procedure that uses such a hierarchy in computer system design is developed. The procedure uses the hierarchy to trade off cost and accuracy of system performance predictions. The viability and usefulness of the procedure are demonstrated by applying it to the optimization of the architecture of a complex computer system - the CPU-memory subsystem of the IBM System 360/Model 91. In most of the experiments conducted, the procedure converged, if not to the exact optimum, at least to within a very near region of the optimum. A sensitivity analysis procedure is then used to identify the exact optimum, as well as to determine the sensitivity of the objective function to changes in the system parameters. The efficiency of the overall procedure is shown to be considerably greater than that of the worst-case approach to system design.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1977
- Accession Number
- ADA057646
Entities
People
- Balasubramanian Kumar
Organizations
- University of Illinois Urbana–Champaign