Happy Coral Anniversary, System/36 And System/38!
June 19, 2023 Timothy Prickett Morgan
The AS/400 is the product of the marriage of the System/36 and the System/38 that had its first child delivered on June 21, 1988. So, to one way of thinking about it, the AS/400 is not just celebrating its 35th birthday this week, but the System/36 and the System/38 are celebrating their Coral Anniversary. The rules of etiquette say that we have to send IBM Rochester something carved in jade to celebrate.
The AS/400, which lives on as the IBM i platform running on Power10 hardware, is neither stuck on a coral reef nor jaded, but just going on about its business of integrating new stuff and evolving to meet the needs of conservative yet innovative companies who have to actually solve problems and not just buy new hardware and software to keep themselves busy. They’re already business enough, thank you.
It has been an amazing run since “Project Silverlake” was formed out of the ashes of the failed “Fort Knox” project to converge IBM’s five disparate midrange system product lines, an effort that collapsed in the mid-1980s under its own weight and that left IBM Rochester in the lurch with no next-gen product just as Digital Equipment Corp and Hewlett Packard were very aggressive with their proprietary minicomputers and as the first Unix systems from Sun Microsystems, Data General, Hewlett Packard, and not IBM were entering the scene.
It was a terrible time, and is well documented in a book that I randomly found and read as I was researching the development of the Silverlake system, which was launched as the Application System/400 that The Four Hundred has served, in many of its own incarnations, for the past 34 years as a publication. That book, called The Silverlake Project, was written in 1992 by Roy Bauer, Emilio Collar, and Victor Tang, all of whom worked at IBM Rochester and were among the many instrumental executives orchestrating the mad-dash development of the AS/400 in a mere 28 months from the wreckage of Fort Knox and the inspiration of the System/38 and the ease-of-use of the System/36. The book is fascinating, and it tells a story that I have personally never appreciated.
I always knew that the architecture of the AS/400 and its underlying system hardware were unique and brilliant. What I did not fully appreciate were the arduous conditions under which it was compelled into being – just as the System/36 was being made a machine for “office automation” and the System/38 was deemed non-strategic. Tom Furey was appointed director of the Rochester Lab in March 1986 and the clock started ticking to June 1988. They had 28 months to get a new machine out the door, and 2,500 engineers and programmers to create a new system based on the concepts of the System/38, which was more incredibly more expensive per unit of throughput than an IBM mainframe, and yet being able to run RPG II applications from the System/36, RPG III applications from the System/38, and new native mode RPG IV applications in AS/400 mode. Fortress Rochester had to create its own Fort Knox, as it were.
Under the excellent leadership of an outsider named Tom Furey, who was a hotshot back at IBM HQ banished to the prairie cornfields of Minnesota to fix that IBM midrange business, the Silverlake team created whole new ways of designing, testing, and building new systems, and in a very real way showed off the very best that Big Blue could be – and we might say has ever been. There are so many neat things about this story that you really need to read the book if you are an AS/400 enthusiast. But there are a few things that stood out to me.
Because I am a bit crazy about high performance computing as well as the AS/400 and other Power Systems iron, I read with great pleasure that IBM Rochester made use of a gate array simulator supercomputer called the Engineering Verification Engine, created using custom chip logic and predating the commercial launch of FPGAs by Xilinx and Altera, to not only simulate and test the Silverlake processors, but the entire system even before both were manufactured. This enabled IBM to not just build one or two prototypes to test, but thousands of them, with very few defects, which put Silverlake machines in the hands of software partners and customers a year or more ahead of launch, which is how over 2,500 applications were running on the AS/400 on launch day.
The whole Silverlake project shifted everything to the left, and brought everyone – hardware designers, compiler and database software developers, application software developers, marketing and sales people – into the design and testing of the machine, thereby enhancing the design and also making it more probable that the project would actually succeed. It was not a foregone conclusion that it would, and even Furey thought the odds of success were only about 1 in 1,000. Furey was upfront about that with the Rochester team not to demoralize them – Fort Knox weighed heavily on everyone in the IBM midrange – but to motivate them, and his team created a means of showing how those odds kept coming down as they invented whole new ways of doing things.
Yes, the AS/400 might have been a gussied up System/38 that could run emulated System/36 applications, but the whole process of making and selling that machine was tuned up like a hot rod. Like a rocket. And the reason why we are all celebrating this week is because of the mother of invention necessitation that Fortress Rochester embraced as they let down the drawbridges and brought supply chain partners, software partners, and customers in on Silverlake, thereby making sure they could not just do IBM’s first global launch, but its first global rollout of a modern computing system.
We get on IBM a lot about its lack of marketing. But let me tell you, back in 1986, 1987, and 1988, these people were inventing whole new ways of segmenting their many markets around the world and hitting their targets, which is what made the AS/400 the fastest growing business in its long history. IBM sold more AS/400s in the first eighteen months of their availability than it did the IBM PC, which launched in 1981 to a much broader market. Think about that for a second. And the reason why is that IBM Rochester didn’t just invent a new machine, like the System/36 or the System/38, and throw it over the wall to the salespeople and out in the world to see what customers might do with its gee-whizardry. IBM Rochester created a machine that did the things they wanted and needed it to do: Run your business, not your IT department.
Silverlake was an amazing accomplishment, and it still is, and its success centers on what IBM Rochester called the Ten Silverlake Principles:
- Appoint a leader with vision.
- Institutionalize that vision by picking the right people and giving them the right mission.
- Empower your people.
- Use cross-functional work teams.
- Segment your market, choose the appropriate segments, and position your product within those segments.
- Research and model your markets and business.
- Allocate resources by setting priorities.
- Break time barriers by using parallel processes and getting it done right the first time.
- Form partnerships with outsiders, especially customers.
- Shape and continuously exceed your customers’ expectations.
Markets were unforgiving back then, just as they are now. That latter one, maybe just maybe, with Power10, we are seeing a bit more of that. We remain hopeful about Power11 and Power12.
Happy anniversary, System/38 and System/36. You made a fine baby, one that was born in Rochester, Minnesota, on a hot solstice day, and is still a good kid to this day.