The vulnerability of computers: Malfunctions may be due to ¿illegal¿ operations, to hardware failures, or to combinations of hardware and software failures that simply elude pinpointing

Digital computers pose unique reliability problems. Serious information-processing errors can occur through one failure lasting a billionth of a second in one of the hundreds of thousands of digital components that switch billions of times a day. Just in the transfer of data between Computers and their peripherals, a million characters may stream toward a printer within 15 minutes, and millions of characters per second flow to and from magnetic-disk memory units. What´s more, the tasks that computers generally perform are so complex that some errors can go undetected for long periods of time, until outputs from the machine become sufficiently unreasonable to arouse a user´s suspicions. The only way to design reliability into this environment (besides the method of worst-case design that is used in all walks of engineering) is to build various methods into the machine for detecting and correcting malfunctions. Modern computers therefore have complex watchdog hardware and firmware and programs to keep the machines serviceable despite intermittent or continuous hardware failures.