The AS/400 Turns 27, And Still Has Much To Teach IT
June 22, 2015 Timothy Prickett Morgan
The collective brain trust here at IT Jungle expends a lot of its intellectual and emotional energy on watching the ecosystem of customers who use the progeny of the venerable AS/400. Many of you have been using IBM midrange systems since the days of the launch of the System/38 back in the late 1970s and you have been long-time users of the systems for nearly a decade when the AS/400 made its debut on June 21, 1988, and when this platform made good on the name International Business Machines.
You will have to forgive me for still being in college on that fateful day, and for taking a year before I finally got the job as one of the founding editors of The Four Hundred way back when. At the time, I knew a little bit about computers, but I didn’t know enough to be useful. I certainly didn’t know jack about business and how computers were actually used in these things called glass houses. I knew a thing or two about technology, and I like all kinds of technical material and have a knack for it I guess. But the thing that has been truly great about being the editor of this newsletter for so long is the wide and deep and very human education that I have been given by understanding this machine and its follow-ons, and building relationships with the people who have made it more than just a machine, but a platform on which to base a business that makes things or does things and gives people sustenance.
These are, in fact, the important things, and it bears remembering as the AS/400, as embodied in the current Power Systems-IBM i platform or what my colleague Alex Woodie likes to call Power i, was founded on the principle that you build a good machine that lasts a long time in the field and runs and supports applications for a very long time. You build a machine that abstracts away the complexity of a system as much as possible and makes it easier for smart people to build applications that run businesses in a smart way. The IBM innovation was to build a system that makes it possible for other people to do their innovation and not get mired in the muck of information technology. The kind of stuff that I love, by the way.
You don’t know this, but in my other job as co-editor of The Platform, I must remind people several times a week when they come up with some new technology or approach to solving some problem that I have heard of this before. Most recently, I was talking to the high-end switch people at Cisco Systems about their implementation of software-defined networks and how they were abstracting network configuration and focusing it on applications, and I quipped that the Application Centric Infrastructure (ACI) that Cisco was espousing was not a new idea at all, but was embodied in this thing called the AS/400. (Many Cisco folks hail from the former Sun Microsystems, and we had a good laugh about how old ideas are new again.)
It happens time and time again, and as I have written in these pages, many of the ideas embodied in enterprise, hyperscale, and high performance computing were commercialized in the AS/400 and its successors. Mark Funk, an ex-IBMer who contributes to The Platform every once in a while, reminded me that the kind of caching that we need in supercomputers has existed for decades in the System/38 and AS/400. We get it. The world doesn’t, but I am always happy to remind them. Many of the in-memory database innovations, like encoded vector indices, are old hat to us. In fact, what does an in-memory database mean in a world that doesn’t have single-level storage? It means precisely that: Your machine’s architecture was so short-sighted that it didn’t have single level storage at all!
And I don’t think there is anything old or old fashioned about that. And if IBM had any freaking spine at all it would have never changed the name of the System/360 and the AS/400, ever. Marketing for the sake of marketing, to make something appear new that is just evolving, is the stupidest thing in the world. IBM does it again and again and again, convincing itself that this will somehow magically help. It helps Olgilvy & Mather to big bags of money, to be sure.
Do you honestly think that Amazon Web Services will ever be dumb enough to change its brand? Or Google Compute Engine? Microsoft had to get Windows out of the Windows Azure name because Azure also supports Linux, and so its cloud has to be called Azure (alone) if Microsoft wants to build a public cloud that is open. All of this talk of branding just sounds so utterly ridiculous to me.
I am a big fan of Henry David Thoreau, as many of you know. And some old wisdom bears repeating.
“Every generation laughs at the old fashions, but follows religiously the new,” Thoreau wisely said, and at the end of Walden there is this gem, which I look at time and time and time again:
“However mean your life is, meet and live it; do not shun it and call it hard names. It is not so bad as you are. It looks poorest when you are richest. The fault-finder will find faults even in paradise. Love your life, poor as it is. You may perhaps have some pleasant, thrilling, glorious hours, even in a poorhouse. The setting sun is reflected from the windows of the almshouse as brightly as from the rich man’s abode; the snow melts before its doors as early in the spring. Cultivate property like a garden herb, like sage. Do not trouble yourself much to get new things, whether clothes or friends. Turn the old; return to them. Things do not change; we change. Sell your clothes and keep your thoughts. . . Superfluous wealth can buy superfluities only. Money is not required to buy one necessary of the soul.“
Turn the old; return to them.
Here is what IBM forgot since the 1990s, since it got spooked by a near death experience and the rise of so many rivals. The real thing to do is to build the best product you can, better than the competition, and sell it at a reasonable price to help people solve real problems. An IBM scared by the Internet and quick rivals was constantly engineering its financial results with obscene levels of stock buybacks. This made the top brass at Big Blue and investors rich but sold out the AS/400 customer base to a certain extent to win a war against Sun Microsystems and Hewlett-Packard a decade ago–I could go on and on. IBM took its eye off the ball at this time and lost the real war in the midrange to Microsoft Windows Server and then Linux came in and made it even tougher.
The AS/400 was the best damned machine IBM ever made, and I suspect ever will make. I sometimes think the real story of the AS/400 is that it was a freak accident that could never be repeated. But the VAX from Digital Equipment was, in its own way, an equally impressive and elegant machine, aimed mostly at technical workloads.
All that I ever hoped was for it truly to go mainstream, for it to be normal, not weird. That would have required IBM to want to have millions or tens of millions of customers instead of hundreds of thousands, and to extract the same profit pool from a much larger customer base. This clearly did not suit Big Blue, and the managers who did not understand what they had and what they could make of it.
So as I personally celebrate the AS/400’s 27th birthday, here is what I wish. I wish IBM would start an effort to make a new AS/400, one that has Linux commands but with the original architecture that IBM started out with way back in 1988.
Remember when the database was the file system? Something that Microsoft tried to do with WinFS and failed miserably at a decade ago? Back when, in OS/400, there was no separation between objects of any kind stored in the database, the kind of thing that academics are still wrestling with and think has not been invented–much less used in production by midrange shops before IBM freaked out and tried to turn OS/400 into Unix and then sold out OS/400 to Unix. I never had a problem with adding things to OS/400. But taking away its uniqueness and not making it a mainstream volume product, I had a problem with that and I still do. I get a big kick out of the hodge-podge of Hadoopery to try to slap together something that can handle unstructured and structured data and run SQL queries against it, with such complexity and such an enormous cost in compute power, that it is laughable.
We have such cheap computing these days that if you could port the old database file system from early OS/400 and go back to its intellectual roots with a hybrid OS/400 Linux, you might even be able to revitalize a real Power Systems business that could take on Intel. While I am glad for all of our sakes that IBM is opening up the Power chip architecture so others can build systems, IBM i is not going to be on these cheaper, scale-out systems. So my enthusiasm on your behalf is somewhat muted. Sure, Google is going to get some inexpensive two-socket servers for running its workloads, and so will a whole bunch of companies running Linux in hyperscalers in China and North America, but you are not going to get them.
If I can’t have IBM i go hyperscale, then let’s build Silverlinux for a new Silverlake. Take all the lessons we learned and rebuild the Application System/400 and at the same time build a better Linux. We can make it emulate RPG and COBOL programs and skin the DB2 for i database for all I care. We can skin OS/400 commands on top of Linux commands. Whatever. IBM can open source the pieces and reinvigorate this business, and mix it up with OpenStack and BlueMix and whatever else it wants. IBM understood that Linux was important back in the late 1990s, but it did not understand that it, not the Linux collective, had a better system architecture.
To hell with celebrating a birthday, and to hell with arguments about names. I am over that. I want a rebirth that brings tectonic shifts to the midrange. I know this is crazy, but I don’t care. We need some system architects and source code to start. We need some people who probably don’t want to work anymore, or heaven help us, are no longer with us. Anyone want to play?