Architecting for failure: how to minimize the returns on home networking and automotive markets

During presentations at the GlobalPress Forum this week, two items of information caught my attention as both important and related to each other. One emerged during a refreshingly candid presentation by Broadcom CTO Henry Samueli. He showed an actual case study of an attempt to create a fully networked home. The result was a Babel of incompatible networks, incompatible protocols, and literally a closet full of internetworking hardware.

I might have dismissed this as an isolated misfortune, except that almost the same scenario reappeared later in the conference, this time in an entirely different context. In a panel on automotive electronics, one panelist observed that a 2008 high-end sedan could easily have over a hundred microcontrollers and microprocessors, interconnected on a dozen different kinds of busses, with the whole system partitioned into regions, each region with its own concepts of simultaneity, allowable latency, data formats, reliability, and security. Many of the processing sites will come from different vendors working with different development tools.

It hardly needs saying that this is no way to do system architecture. The probability that the resulting system—in either the home network or the automotive case–will function reliably is very small. The probability that the system will approach optimum cost, performance or energy efficiency is zero. While this may be an annoyance in the networked home—I can live with the lights in the bedroom getting disconnected from their switch every time the refrigerator runs low on milk—it has more serious implications for a moving vehicle.

But the problem also has implications for the vendors who are rushing to make their fortunes in these markets. Historically, markets have tended to develop fastest when architectural standards ensured a coherent, reliable system architecture, independent of particular vendor biases, from the outset. Conversely, in markets where a significant degree of coupling exists between subsystems with poorly-defined interfaces, the result is almost always poor product performance, disenchanted customers and unacceptable return on investment for vendors. Compare, if you will, the world of GSM phones to the world of video conferencing.

All of this brings to mind another random remark during the conference, this time from one of the European journalists. He said he had been discussing the tangle of competing broadband wireless standards with an industry executive recently, and the executive commented that, in retrospect, the development of the GSM telephony standard had been a remarkable achievement. The executive didn’t believe that the social order or the will any longer existed to achieve a single, comprehensive standard of that quality across the communications industry.

That may be more than a social observation. It may be the epitaph for the unregulated, disaggregated global communications industry we have created.