A Heuristic Integer Programming Approach to Certain Types of Computer System Design Trade-off Decisions

摘要

The work uses some modified integer programming techniques to attack certain types of decision problems arising in computer system design. These occur when an alternative selection is to be made over implementations available for several subtasks making up some larger task. The initial model investigated is an integer constrained optimization problem with objective and constraints expressed individually in inner product form and integer (1,0) variables used to express selection or non-selection. Objective and constraint attributes can consist of serial execution time, nonrecurring and recurring costs that are essentially additive in nature such as man-months, manufacturing costs, weight and others. Constraint bounds can consist of maximum usage of these resource types including interrupt time restrictions. It is pointed out that for many problems the cost of obtaining reliable data for implementations with respect to the preceding characteristics can exceed the benefits obtained from good use of them unless the problem addressed is relatively well known. The constrained optimization model is shown equivalent to the multiple choice problem appearing in the operations research literature. (Author)