AN EXPERT MODEL BASED ON USER/PROCESS PROFILES FOR COMPUTER SYSTEMS EVALUATION.

Abstract

This dissertation describes the development and implementation of a rule based expert system for the design of computer systems' hardware configurations utilizing a portable workload model based on user/process profiles, local optimization rules and architectural constraints. The systems' configuration expert was designed within the framework of an expert system lifecycle model. The lifecycle model provides an useful methodology to support the design of expert systems. Expert systems are proposed as a means for automated support for decision making in systems analysis and design, especially for those less structured tasks that occur during problem definition, requirements analysis and logical design. The development of the systems' configuration expert was simplified by the use of an integrated rule base management tool: NEWES. NEWES allows for data driven inference, flexible conflict resolution, multiple concurrent rule bases, relational data base like access to short term memory and frame based object definition of short term memory.