In the early days of computing, it was much easier to identify and fix whatever was causing a processing slowdown. Today, though, that approach seems a whole lot harder to implement About 40 years ago, I was a programmer for a municipality. We had a mainframe but the bulk of the programs simply duplicated the card readers, sorters, and printers. The computer was slow to a degree unimaginable to most programmers today. One program that had actually been written for the computer was taking 45 minutes to run (we had no idea of what multi-processing was) and I was assigned to make it run faster. In researching the problem, I discovered that there was a brute-force table lookup of nearly 1,000 items. There was no such thing as indexed files. I introduced the concept of a binary search (I didn’t know there was a name for it) using Cobol coding. That sped up the processing such that the program ran in 20 minutes. In this instance, I optimized the slowest thing and it was eminently successful. Later, without the knowledge or approval of IS management, I wrote a payroll processing program that chopped out the entire manual card sorting operation then in use. Our payroll system till then was a card-based system with about 20,000 cards. Card sorting took from a day and a half to two 24-hour days, very labor intensive and prone to inaccuracies. They then read the sorted cards into the computer and printed a preliminary register. Payroll made corrections, cards were punched and manually placed in the trays, and it was all read over again to print the final payroll register and once more to print the checks. My program let them keep the cards (selling a computer tape file concept was a little much at the time.) I simply accepted the cards from their natural filing order and sorted them on the computer, which was not something management had seriously considered. The computer sorted the whole thing accurately and allowed updates from payroll for corrections and was ready to print registers or checks when Payroll was happy. This was implemented by the operations manager (who was covertly cheering me on) as a test and when he saw two days of card sorting disappear it was put into operation immediately. Our payroll processing time was cut by two to three days. I did not get fired for smuggling my ideas in on City time, but management didn’t much appreciate being ignored. The Payroll section was really happy, as was our operations manager. A computer-based payroll system was not far behind. Again, optimization of the slowest thing. Yeah, that kind of behavior is totally impossible now – but there was a time … Data Management