This paper deals with the use of mathematical modeling in analyzing the economics of mainframes and superminicomputers. It challenges many conventional models, which have been based on Grosch’s statement that “to do a calculation ten times as cheaply, you must do it one hundred times as fast” [1].
In this eight-page paper, the author first examines problems with previous studies relating computer cost and power--eleven such studies are summarized. Having shown some pitfalls in these studies, he then proceeds to review the theoretical basis used in an alternative approach and the computer characteristics that can be considered in the analysis. The computer characteristics chosen are millions of instructions per second (MIPS), main memory size, number of input/output (I/O) channels, and computer power. Computer power represents relative performance based on a rating system developed by International Data Corp. Cost data considered in the study include both purchase and lease prices.
The final results, based on 359 computer systems surveyed, are interesting but not surprising: newer systems cost less and have higher relative performance than older models. Also, IBM-compatible systems cost more and have lower relative performance than non–IBM-compatibles. And finally, larger computers do not necessarily imply lower computing cost. Consequently, the author concludes that the current trend of decentralizing computer resources rather than acquiring very large systems makes economic sense.
This paper would interest people concerned with the mathematical modeling of computer performance and cost.