There was a time companies selling into the aviation and defense markets had dedicated mil/aero business units. Today, increasingly, suppliers are setting their caps at a much broader target—high-reliability components and systems. As intelligence and functionality increase in application areas like medical, transportation, communications, infrastructure, and industrial, the price of failure has become unacceptably high. As a source involved in chemical processing commented, "In our industry, we don't make mistakes, we make craters." In cases like these, the reliability of the components becomes essential.
That's a difficult challenge to meet in an era when shrinking critical dimensions have made even terrestrial components vulnerable to single-event upsets, a phenomenon that was formerly restricted to space electronics. Device designers find themselves pressured to provide top performance, developing highly-dense chips that dissipate more heat, which can, in turn, reduce lifetimes. Amid these realities, the requirements for hi-rel industries are nudging toward the traditional needs of the mil/aero market.
MIL-STD-721c, a U.S. military document developed to provide a common language for discussions of reliability, defines the concept as follows:
1) The duration or probability of failure-free performance under stated conditions 2) The probability that an item can perform its intended function for a specified interval under stated conditions
Since the mid-1990s, reliability engineering has been driven by the physics of failure, which centers about determining failure modes and then designing to correct them to a sufficient degree to achieve that specified interval.
In some ways, the suite of hi-rel applications has become so broad that it begs the question where do mil/aero applications leave off and other high-reliability applications begin? The query is particularly relevant given the increasing focus on the use of commercial off-the-shelf (COTS) components. A vendor can design high-reliability components based on physics-of-failure concepts, but if the "specified interval " falls below that required by mil/aero applications, the devices won't serve the purpose.
How do you feel about the high-reliability offerings currently on the market? Do they provide the performance and lifetimes that you need for your systems or has "high reliability" become a marketing phrase in the same way "light" works in on food indu packaging stry and "green" works, well, everywhere? When you buy hi-rel components, do they generally meet military and avionics standards or do you find yourself doing signification testing and development on your own? And, most important, how do you define high reliability?
Actually, the expectation of high reliability has increased. As more things go electronic, the users expect the devices to work first time, every time. The fastest way to get your product panned in the media is to have someone tell someone else that your stuff is crap. You are done, no appeal, no pardon, that's it.
So if you do not design your device to be reliable, you will soon be out of business.
There is no one-size fits all answer.
High reliability clearly means different things in different markets, as does product failure.
For example, the last paragraph of this article has an error that was missed during editing. Does that matter? Not really.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.