IBM i at 35: A Walk Down Memory Lane
April 26, 2023 Alex Woodie
You may have heard that the IBM midrange platform is turning 35 this year. IBM has a number of events planned in honor of that milestone, culminating with a big birthday bash on June 21. IBM execs gave us a sneak peak of the festivities to come at this week’s POWERUp show in Denver, as well as a technical look back at exactly how we got here.
The day the AS/400 launched in 1988 was notable for several reasons. IBM i CTO Steve Will, who was just starting his distinguished engineering career at IBM, recalls lots of excitement in the air.
“We did have several of the actors who played M*A*S*H characters there with the announcement,” Will said during the “IBM i @ 35” keynote at the COMMON POWERUp 2023 conference on Monday. “It was a fun day all those many years ago, in spite of it being as hot as I’ve ever been in my life, and I’ve been by a volcano.”
IBM hired Alan Alda to star in a TV ad about this new thing called the AS/400. Alda was there on the Summer Solstice in 1988, which turned out to be an intensely hot day. How hot was it outside of the IBM lab in Rochester, Minnesota?
“It was so hot on the tarmac,” said Alison Butterill, the IBM i product manager, “that those chairs with the metal legs were sinking down into the tarmac, and periodically you could see people [fall over] as their chair sunk too far, or one leg did.” There’s no word if any Hollywood actors met an asphalt fate that day.
The AS/400 was an immediate hit, and soon the big white boxes were flying off the Rochester assembly line. Demand for the AS/400 was as hot as that launch day in Rochester, as companies couldn’t wait to move up from S/36s and S/38s to the Next Big Thing in business computing.
So many businesses immediately ordered an AS/400 that many of them apparently received their new server at the same exact time. “I can’t tell you how many times I’ve met somebody who had the experience of having the very first AS/400 delivered to them,” Will said. “There are 10 of you in the audience who will claim that you had the very first AS/400. But in any case, I love hearing those stories.”
There are so many stories. Malcolm Haines, IBM’s marketer extraordinaire, documented several of them nearly 20 years ago as part of his “iSeries.mySeries” campaign. Some of them took on the air of urban legends as part of Haines’ “Legends of iSeries” series that was associated with that campaign, and which lived on the iSeries Nation website. (Anybody remember iSeries Nation?) Will and Butterill shared one of those videos during their keynote on Monday – the one about the server continuing to run despite being locked away in a 140-degree closet for two years – and hopefully there will be more as we get closer to the big day.
But the evolution of the AS/400 into today’s modern IBM i server wasn’t all fun and games. While the AS/400 was advanced for its time – thanks to features like single-level storage, an object-oriented architecture, and a technology independent machine interface – the IBM i has evolved far beyond the original vision of the AS/400’s creators. There were some stressful moments for the Rochester crew, and perhaps a little uncertainty, as the box evolved over three-and-a-half decades.
For example, the original AS/400 was based on a 48-bit Complex Instruction Set Computing (CISC) processor. But IBM Rochester couldn’t wait to get away from the added complexity that the CISC chips entailed, Will said.
“We had to get to a Reduced Instruction Set Computing platform, the RISC platform,” he said. “But that meant that none of the code that was below the machine interface was any good anymore. We had to rewrite everything we had in order to do that.”
IBM had to rip out everything below TIMI and rewrite it. Back in then (around 1990), just to add to the confusion, there was vertical machine code and vertical machine code in the AS/400’s guts. When it was all said and done, the machine code was dramatically simplified and there was no longer vertical and horizontal code, but it wasn’t an easy task.
“That effort was hundreds of people working almost around the clock, almost seven days a week for a long time to make sure that that new release could come out and you guys could benefit from that,” Will said. “It was a very strenuous time and we were all very happy when we finally completed and could move on from there.”
While some customers reported problems with their application code, the migration to 64-bit RISC processors went about as smoothly as one could expect, and the vast majority of customers’ code kept running as it had before. “This is where I get to remind people that the architecture really did protect us as we moved forward in technology,” Will said.
The end of the decade brought another big moment in the evolution of the box. With the launch of OS/400 V4R4, IBM unveiled the Portable Application Solutions Environment (PASE), which allowed the AS/400 to run AIX workloads. The introduction of PASE was a milestone in the evolution of IBM i, Will said.
“So much of what we were able to do today, so much of what we evolved into, was because we were able to marry into the ILE based stuff that compiled to the machine interface with the stuff that runs in an AIX-like environment in the same processor, [with] the same control,” he said. “It’s critical. For example, all of our SAP customers are running in that environment, and we’ve got 1,000 SAP customers, so it’s pretty important.”
The IBM i server is known for having an integrated Db2 database. But for a long time, most IBM midrange customers didn’t use the database to its full potential, and didn’t use the language most associated with relational databases, SQL. That started to change about 15 years ago, when Dave Nelson, who today is the director of development for IBM i, got serious about moving to SQL.
“The initial internals of the database was built sort of like a simple engine,” Will said of the Classic Query Engine (CQE). “You could ask a question and get an answer out of it. It couldn’t be a very complicated question and it would take a little while before the answer would come out.”
As customers started storing more data in the IBM midrange platform and started asking more complicated question, it soon became apparent to the leaders in Rochester that the CQE wasn’t going to cut it.
“We were going to have reinvent that,” said Will, who took over what today is the IBM i Architect job in the fall of 2005. “So we created a new query engine, the SQL Query Engine.”
The new SQE, as it was called, would execute SQL queries atop SQL tables, as opposed to the CQE, which worked directly atop DDS entries. According to Will, developing the SQE into the sports car that it is today took a bit of work, and getting people to use it wasn’t always easy.
“It took us release after release after release of making that SQL Query Engine the very best it could be to compete with any other database out there in the marketplace,” he said. “And yet we were still never able to completely get rid of the classical query engine because there are still queries that run better in that path than with the sports car.”
Similarly, the introduction of multi-threading also marked a departure from the ways of the AS/400 past and into a bright and glorious (and fast) future.
“Through the years, the whole industry changed,” Will said. “When we created the S/38 and AS/400, there was no concept of threading applications. But as the rest of the industry had to deal with processors that were not capable of doing quite as much, they started getting into threading as a way to do applications. It turns out to be a really good thing to do for many styles of applications. And while we avoided it for a while, resisted it for a while, ultimately we had to solve the problem with … a good strong multi-threading environment.”
Lotus Notes provided the testbed for multi-threading on the AS/400. It worked so well that IBM ran its entire email system on the AS/400 for years. And when the Internet revolution happened, the investment in multi-threading really started to pay off.
“Because we were going to start doing multi-threaded things, such as Lotus Notes, we could now handle the multi-threaded applications that were coming in across the Web,” Will said.
The Web marked a big moment in the evolution of the architecture’s single-level storage environment. While it was innovative and paid dividends for traditional business applications, single-level store enacted a performance tax on data that made its way through a Web server. IBM knew it had to deal with that performance hit without compromising other parts of the system.
“So we had to create a parallel storage model, which we called Teraspace, which allowed us to use the same kind of pointer lengths, but be a completely different address space,” Will said. “So we had to change the architecture so it was not just single level storage, but it was this pairing of single level storage and Teraspace.”
Will had just taken over the chief architect job just before the iSeries and pSeries convergence in April 2008, which resulted in IBM i and AIX running on Power. He was still new on the job when he was tasked with making the big announcement for the COMMON audience in Nashville.
“I did not make the decision to converge the platform, and I certainly wasn’t involved in the naming of the platform i,” he said. “But I got to stand up in front of you and say ‘The name of the platform is IBM i!’ and just be sure of that. That was really fun.”
The next big moment was the launch of IBM i 7.1 in April 2010. This release was important for a few reasons. First it started the beginning of IBM’s Tech Refreshes (TRs), which it continues to this day. Secondly, it marked the first time that two different IBM i operating systems could live on the same box.
“That dramatically changed how we did things as basic as upgrade from release to release,” Butterill said. “You would fire up a new partition in a new release, 7.1 for example and you would still be hosing the 6.1 release that was continuing to run your business.”
Lastly, IBM i 7.1 marked the release of Free Form RPG, which has helped to modernize the language and make it more attractive to a new audience of developers. The launch of Free Form RPG was also important because of the way IBM came to the decision to implement it. The company had relied on various advisory councils, including COMMON Americas, COMMON Europe, the Large User Group, an ISV advisory council, and a partner group. But they couldn’t agree on what were the most important new features. With 7.1, however, consensus finally came together that Free From RPG was important, which gave IBM the confidence to implement it quickly.
“I can’t necessarily say we pushed our advisory councils to push this to the top,” Will said. “But we were certainly very happy when they did finally decide that was the most important thing. Because you will see . . . a lot of what we’ve done over the course of the last 10 to 15 years is openness and the ability to bring people from other backgrounds into IBM i is absolutely critical to the future of this platform.”
There is one last important event in the development of the platform that Will and Butterill mentioned during their keynote. The biggest IBM i customers had long had a requirement to be able to upgrade without downtime. However, implementing the type of clustering needed to meet that requirement was complicated by the presence of IBM i’s single-level storage architecture.
“That was a requirement that we weren’t sure we would ever be able to find a way to deal with,” Will said. “But converged into that requirement were many, many, many financial institutions which began to say, look, we can’t be financial institutions if we don’t have a way to stay continuously available. If you want to remain the core of our banking solution, you have to find a way to be continuously available.”
The Rochester brain trust came together and looked for a solution that could meet the requirement without violating the precepts of single-level storage. It’s something that had stumped the best of Rochester for decades, Will said.
“But smart people – Mark Anderson, who was the database architect forever and ever on this platform and Kris Whitney and some smart folks on the team – they found a way to do this so that we could not only remain viable but be extremely agile in the kind of things we were going to do,” Will said. Db2 Mirror, which was delivered with IBM i 7.4.
All of these changes – going from CISC to RISC, the introduction of the SQE and multi-threading, the convergence of platforms, Free Form RPG, Db2 Mirror – are examples of how the platform has moved forward technologically but without moving away from its roots.
“It’s much more complex than it was in the past,” Will said. “It is the same architecture, and it’s not the same architecture. It has the same core tenets. In fact, it has many of the same basic principles that we’ve always had.”
IBM Brings Active-Active Mirroring Into Db2 For i Database
IBM Officially Announces i/OS 7.1
It’s Official: Now We’re Power Systems and i for Business
Thanks for the history lesson — I lived through all of it. But you may want to edit this sentence:
“Back in then (around 1990), just to add to the confusion, there was vertical machine code and vertical machine code in the AS/400’s guts.”
IMHO: it is one of the best server architecture we have today. To protect investment and compatibility in business logic is a immense value for a company and it is a spirit that must be kept as a core value (“…really did protect us as we moved forward in technology…”… elegant architecture pays off in the long term, in this market blinded by short term goals… nothing great was built in the short term).
But enough auto celebration: one of the big mistake was not updating the 5250 protocol, nothing fancy, some fast grid component, big screen use, date pickers, something geared to build ERP not games…. With an improved good graphic, conversational, terminal protocol (not web!……) this machine could have been unbeatable in dev time, op time and final results, feasible today with all the tech we have and (archived / open source) projects we can tap into / reuse to implement it and I fail to see why this is not seen as a big priority (even 20 years ago…)…. the interface is usually the glasses through which the system is perceived by laymans…
No mention of Frank Soltis?