WorldWide Drilling Resource

Partial Failure = Total Failure by Britt Storkson Owner, P2FlowLLC I was speaking with a farmer the other day, and he was telling me about an automated crop sprayer for which he had purchased a software upgrade. He said it fixed problems in one area, but introduced problems and inaccuracies on other areas. In other words, it worked, but it didn’t work well. I’ve observed this phenomenon over and over again in industry where a software product would perform correctly most of the time, but would malfunction some of the time. The time the software malfunctioned could be just a few seconds, but when human lives depend on the decisions a computer makes - such as flying an aircraft - those few seconds can mean the difference between an uneventful flight and a multiple fatality incident. This is how we get cars that accelerate out of control for no apparent reason, or transmissions which refuse to upshift (like my father-in-law’s minivan does from time to time). If we are dealing with a cooling pump that stays on too long with the only harm done is wasted energy, it can be tolerated. When human lives depend on the computer software, this software must work the first time - every time. Any failure - even for a few seconds - cannot be tolerated. Often, the people who are tasked with making this software work - the operators - have to “cover” for its flaws. They must remember to avoid certain entries or not do certain things to cause the computer to malfunction, or reset it when it does. While most operators (including airplane pilots) are very competent, but taking into account the complex, changing nature of the typical modern computer system, no one can remember everything. This begs the question: Why does the operator have to “cover” for the computer system in the first place? How did we get here? Two things: 1. Computer overcomplexity, driven primarily by computer vendors who “pile on” fea- tures and functions (complexity) solely to add cost . . . not to make the computer any better or more reliable. 2. Inadequate testing which is made far more difficult by overcomplexity. Companies often do not give product testing the attention (and budget) it deserves because there is no immediate return for the money invested. The company financial analysts cannot quantify what testing should cost and when it will end. It’s be- cause there is no predefined “endgame” for proper testing. Testing is complete when the product works the first time, and every time, under every condition it can be expected to encounter during its lifetime. While many have tried to find a lower cost alternative to product testing, there is none. One must simply live with it over time. If it has a problem, sooner or later the problem will manifest itself. When identified, the problem should be fixed immedi- ately, then retested. There is no other way. Britt Britt Storkson may be contacted via e-mail to michele@worldwidedrillingresource.com 33 WorldWide Drilling Resource ® AUGUST 2019