Embedded systems won’t cause the millennium meltdown many people think, but businesses must ensure that engineers, and not IT professionals, work to make their systems complaint, analysts warned this week. Speaking at a Gartner Group symposium, research director Andy Kyte said embedded systems, found in almost all non-PC devices, do not pose the Armageddon-like threat portrayed in recent press reports. He urged companies who have not yet begun work on their systems to start immediately, by carrying out risk assessments and not to get bogged down in what he called ineffective analysis/paralysis inventory situations. Kyte divided the systems into three groups. The first, microcontrollers, are present in more than 80% of all the Silicon devices on the planet, typically in domestic and consumer electronic products such as washing machines, vacuum cleaners, door chimes and refrigerators. Microcontrollers are non- programmable and are rarely found in combination with a real-time clock. As such, Kyte said there was a one in 100,000 chance that systems containing the controllers would fail. These devices don’t even know what planet they’re on, let alone what day of the week it is. Kyte said. And in a challenge which drew loud laughter from the audience he added: If anyone can find me a fax machine that won’t send or receive faxes over the turn of midnight on January 31, 1999 then I’ll eat it. The second group, microprocessors, are programmable devices and provide a level of sophistication and complexity vastly greater than microcontrollers. On their own, they do not pose a Year 2000 threat but when working in conjunction with a program that has either direct (co-mounted on the motherboard) or indirect (over the network) connection to a real-time clock, they can cause problems. Even still, Kyte predicts that just 7% of microprocessors co-mounted with a real-time clock will demonstrate transient Year 2000 anomalous behavior, and only 2% will show persistent problems. Processors with indirect contact to real time clocks have less than 0.25% of failing, he said. The third category, large scale embedded systems, pose the most serious threat. They are found, alongside microprocessors and controllers, at the heart of many critical safety control systems, including those that monitor manufacturing processes on factory floors; telephone systems; petrochemical plants; traffic light control systems; elevators and so on. Kyte predicts that through 2001, at least 35% of these systems will demonstrate anomalous data processing. For the majority of all embedded systems (in 95% of all cases), the best way to avoid problems over the millennium is to circumvent them, by switching the machine or device off altogether, Kyte said. Where this is not possible, he advised companies to decide what needs to be replaced outright and what can simply be fixed within the time and budgetary constraints. Companies should also develop a resumption plan, should things go wrong, and work out the schedule for weekend working over the millennium as early as possible. He stressed that, unlike the IT department, most embedded system problems won’t manifest themselves until the Year 2000. He said it was critical for companies to ensure that engineers, and not IT professionals, are tasked with carrying out the risk assessment and replace or fix procedures. The most common cause of failure in embedded systems so far is where the ownership of the problem has been put in the wrong hands, Kite said, there are some common points, such as the need to start by raising awareness, but in every other aspect the problems posed are so different that they demand entirely different treatment. Jim Duggan, research director, said he thought telecom systems, petrochemical equipment and power grids would be worse off. He added that aircraft systems would be fine, although their maintenance systems would be susceptible to failure.
