By : Jim Pinto,
By : Jim Pinto,
Dick invited me to be a guest columnist - and I came up with this article. While reading it recently, I was surprised that some of the ideas are still valid today. I've included the article here, to provide a historical perspective on the still important topic of industrial computers, communications, fieldbus and ethernet.
Programmable Controls Magazine, November/December 1987
To speak of many things
Of MAP and DOS and OS/2 of Cabbages and Kings.”
The industrial automation environment is a peculiar battlefield at the present time. Users are caught between a capital-squeeze and demands for improved productivity. Vendors are caught between demands for miraculous results (which they are unable to service) and their own shrinking sales and profits.
One cause of the mess is an excess of emerging technology awaiting proper implementation. The key word here is emerging, because users and vendors alike are distracted by the constant surge of available high-tech options everyone expects and few understand. Many new concepts appear simple. But, while they are elegant abstractions, practical implementation is an arduous and painstaking process which no one seems to expect to undergo.
The history of office automation ties in with the current industry situation. There was a surge of growth in the office environment, which was first dominated by mainframes and then minis and micros, and Apple exploded to the $2 billion sales level in less than a decade. The surge in growth was fueled by poor productivity at lower levels in the office (finger-bashing of manual typewriters and calculators) where there was indeed a vacuum for word-processing and spread-sheets. Suppliers felt that the same revolution was due in the factory. When the office PC explosion slowed a bit, IBM and its horde of hungry hangers-on turned their attention to the industrial environment, expecting similar conquests. So, CIM and MAP are expected to yield miraculous productivity improvements. But where are they?
The hierarchy of the industrial environment extends from the office as shown in the figure. It is easy to imagine each independent unit being controlled by a computer of some kind, and all the units being tied together with a communications linkage throughout the hierarchy. With that, when one machine (or unity) is down, or causes some changes, the entire network reacts instantly, miraculously, with optimized production, and joyous harmony. So, MAP envisages a common communications mechanism which ties the industrial world together.
Unfortunately, implementation is impossible without compromise, and the very compromise bastardizes the idealistic concept. It’s like inventing a universal language: wouldn’t it be wonderful if all human beings could communicate with a common vocabulary? Most humans have a vocabulary of just 1,000 words and so it should be relatively simple to standardize certain sounds. So, why not get an engineering committee to define seven layers of semantic understanding? Unfortunately, the moment agreement is reached in Olympus, competing sub-committees would form to define various dialects and humor variants, which would rapidly fragment standardization and impede progress.
There is a business side to all of this, too. To implement MAP on a PC presently costs about $3,000 to $4,000. People think the problems will disappear when the price comes down. Suppose silicon ships to implement MAP were instantly available as a commodity (subsidized by GM, supposedly) at $1 per piece. IBM, as the leader, would immediately need to introduce its own micro-channel variant of the chip, which it could sell as a non-commodity (with certain performance advantages at a higher price) while the Taiwanese and Koreans would find a way to introduce the commodity at 10 cents a piece.
This is about what happened to DOS and the PC. Apple was in search of a home-computer, and stumbled on the vacuum in office productivity at the lowest level. IBM entered the fray, to fill the need for a big name player with an established office-image, and achieved $2 billion in sales in a couple of years (much to its own surprise) with hardware that was thrown together in haste, and an operating system that was borrowed from a fledgling Microsoft headed up by a brilliant ex-teenager (I suppose we are all ex-teenagers, though not as recently as Bill Gates). With an installed base of $20 billion and 20 million units, the PC became a de facto standard. Microsoft (with the operating software) and Lotus (with the applications software) got to $250 million in sales. But, of course, IBM could not simply continue to sell the commodity it had inadvertently spawned with its open-architecture, and they inevitably came up with micro-channel architecture (to replace the PC-BUS) and OS/2 to replace DOS. Tell me, how else could you continue to make at least 125% after tax against Japanese and Koreans who barely make 2%, if that?
On the industrial side, PC-BUS made an entry against VME (Motorola 68000) and Multibus (Intel 8086) and STE (8-bit), and quickly exceeded all of these with the flood of software and hardware spilling over from the office. Unfortunately, PC-DOS is not multi-tasking, and software developers had to come up with peculiar tricks to overcome the disability. UNIX squeezed down into the PC/XT, and with a little more comfort into the PC/AT-but still primarily through Microsoft with XENIX, which takes almost 10mb on a hard disk and forever to install. After doing that, UNIX is round robin time-sliced multi-tasking and still not real-time. And, or course the predatory marketer IBM could never sink so low as to really support UNIX (which belongs to AT&T). So, OS/2 is targeted to have real-time possibilities, and is certainly not a UNIX derivative, but an intricately manipulated DOS-extension. With Compaq reaching for the $1 billion level, and Tandy and AT&T and the hungry hordes behind, Microsoft must show its independence by offering OS/2 to everyone, while IBM offers its own variant with links to its hierarchy in the upper echelons of the offices of Fortune 1000.
But, isn’t it strange? The 80386 hardware is ready from a dozen sources and even IBM, while OS/2 has yet to be introduced for the 80286 and who knows when the 386 version will emerge?
Meantime, on the hardware side, IBM has abandoned the PC/AT bus and open-architecture to push micro-channel. Hardware developers are invited to design for micro-channel, though they will need an IBM-registered ID-number to be recognized. If one needs more than one serial port on the PS/2 Model 50, one probably gets stuck with a hastily put-together card from AST or some other burned-the-midnight-oil-after-April-2 vendor. Also, the page-scanner for your desk-top publishing system probably cannot be plugged into your PS/2, so you’ll still need an AT.
Getting back to the factory environment, we can still network a PC with a $500 RS-485 network card at 1 Mb and twisted-pair wiring, while waiting for MAP. We still communicate with A-B and Gould PLC’s at the front end, at a snail’s pace over their 9600-baud data highways. With a $1 billion installed base, programmable controllers (PLCs) are still the established front-end real-time-controllers, though they are nervously eyeing encroachment from ruggedized personal-computers like the GE-Fanuc CIMSTAR. Why would you hook-up your I/O to network through a proprietary PLC protocol, and then through a PC, when you can hook your GENIUS I/O directly to a CIMSTAR? The only reason to stay with PLCs is the establishing learning level and commodity pricing. A-B can fight that out with the Japanese. IBM and Action Instruments will look for greener pastures at the next emerging level. Is that cell-control?
In the industrial environment, there are only 3 computer companies in the US-IBM, DEC and HP- and the rest can find their niches. Cell control, if you define it, will be divided up among those three, with DEC probably getting the major share because of its established record in the industrial environment and because of its strong position in networking, top-to-bottom in its architecture. Because of DEC’s strength, it was able to point out that MAP was an emperor’s clothes situation with GM. Now, incredibly, there are talks of including Ethernet somehow into the MAP specification. It’s like writing a new specification for English, to include some of the Japanese alphabet and phraseology.
IBM’s abandonment of the PC-BUS and DOS will inevitably favor VME and DEC, and DEC moves inexorably up with the industrial-VAX and RT-VAX (real-time VAX) and lower and lower priced VAX derivatives. It is clear that, at least in the industrial environment, DEC can choose to bring VAX in at a price that makes it an easy winner at the cell-controller level. If so, can IBM really stay in the industrial environment as a permanent runner-up?
But answer there came none
And this is scarely odd because
They’d eaten every one."
Return to Index of all JimPinto Writings
Return to JimPinto.com HomePage
If you have ideas or suggestions to improve this site, contact: email@example.com