Power Systems i: Thinking Inside the Box
December 7, 2009 Timothy Prickett Morgan
We’re coming to the end of what has probably been one of the longest years in the decade and the beginning of a new decade at the same time. Now is a natural time for us to review the state of the AS/400 platform, now cumbersomely known as the Power Systems server running IBM i 6.1. I am going to take my time and think this through, so I hope you are in no hurry. I have been hurried enough this year for an entire lifetime, and this is a topic that takes some slow thought.
That said, it is difficult to know where to start. Let’s start by being honest. Brutally honest. And for once, let’s actually think inside the box, which is where the applications that matter actually run, after all.
I think the first thing to realize is that things could be worse. Imagine if this newsletter was called The Three Thousand and all of us, seeing the incredible RISC technology that Hewlett-Packard had on deck for its future PA-RISC workstations and servers in the late 1980s, had banked our careers on the MPE operating system, with its own integrated database management system and COBOL applications.
None of us were that stupid, of course, and we bet on the AS/400 horse with the System/36 and System/38 lineage, a horse that eventually won the battle of the midrange that began in the late 1980s and that was pretty much over a decade later when the AS/400 was the only box that was still viable. (Yeah, I know that HP still sells OpenVMS on its Itanium-based Integrity servers, but the word I hear is that OpenVMS development has been shifted to India and is basically in maintenance mode. The DEC VAX has been, for all practical purposes, a dead product line since HP bought Compaq in 2002.)
The fact that the AS/400 has followed a platform convergence path similar to that of the MPE platform–HP moved its MPE environment to its PA-RISC platforms in the early 1990s when the revenue stream for the HP 3000 servers no longer justified unique investments in hardware and software engineering–does not doom the idea of platform convergence. It is how you do the convergence that matters.
Unisys has been gradually moving its MCP and OS2200 mainframe platforms to X86 and then X64 iron–first with MCP as an emulated operating system riding atop Windows, later in a quasi-native mode riding atop what I assume is a hardware emulator. After acquiring Compaq, HP took a few years, but finally finished the job of porting OpenVMS to Itanium, although it was arguably too late to help reinvigorate the follow-ons to the venerable VAX minicomputer. Sun Microsystems‘ efforts for the past few years to foster its Solaris Unix on X86 and X64 iron has resulted in a lot of downloads of free binary distributions of Solaris, but it really hasn’t turned Sun’s fortunes around. I would argue that Sun needed to embrace X86 chips a long, long time ago to compete with the rise of Linux, starting around 1998 when it stepped on Solaris 8 for x86 and jacked up prices on Sparc iron to try to ride out the dot-com bubble collapse by offering worse bang for the buck to legacy Solaris shops on Solaris 9 and its UltraSparc-III iron.
That sound familiar to you? Can you say “AS/400 5250 software tax,” a phrase, by the way, that I coined in the late 1990s when IBM started screwing AS/400 shops over with ridiculously high prices for hardware and software so it could afford to give 50 percent or deeper discounts on RS/6000 iron–which was much cheaper iron than AS/400s to begin with, mind you. My face still gets red with anger thinking about this.
IBM correctly perceived that Sun, HP, and Compaq were threats to its systems hegemony, and ratcheted up the Unix war with ever-cheaper and increasingly more powerful Power-based machines. But somewhere along the way, the AIX crowd in Austin, Texas, forgot that their 64-bit PowerPC 620 and PowerPC 630 chip development efforts failed and that the chip designers in good old Rochester, Minnesota, created the first 64-bit Power chips from IBM–and a damned good design, too. A design, by the way, which is still at the heart of the Power7 coming next year. Ditto for the PowerVM hypervisor, which is a substantially improved version of the logical partitioning that IBM Rochester got out the door in 1998 with OS/400 V4R4. I could go on and on. The point is, the AS/400 used to lead in technology development, and in a lot of midrange accounts, IBM was not embarrassed or ashamed to lead with it.
I haven’t seen that IBM for a long, long time. The IBM I see today–from the outside of course, and it may look a lot different inside–is all cocky that it has won the Unix battle, but no one has the backbone or brains to point out to the top brass that it is losing the midrange war. The IBM I see doesn’t seem to understand what it will take to fight back, to get AS/400 customers to not just spend money, but believe in their own future. I do, by the way, believe that IBM can fight back–and that it must, for its own sake as well as mine and yours. (Just as I believe that Sun should get its act together without Oracle and fight, or that HP should drop Itanium, port HP-UX to X64 chips, and be done with it.) IBM should not be obsessing about how Sun and HP are struggling in the Unix space and how IBM isn’t struggling as much. Goodie for you. Go collect your bonus. But I want a clawback for every AS/400 account those IBM marketeers and sales reps lost, seeing as though we are all one big happy Power Systems family.
IBM needs spend a little less money on share buybacks and a little more money on engineering, sales, and marketing for all four of its midrange platforms: z/OS, i/OS, AIX, and Linux. In short, the answer is to stop being Wall Street’s mistress and start being the midrange systems wife. To be International Business Machines first, and rich later.
In 1998, when the AS/400 business had made it through its transition to 64-bit processors and was doing really well in terms of generating revenues and profits, IBM’s RS/6000 business was still something of a joke, but improving. Flash forward to 2008, and there is no AS/400 division, no one championing the AS/400, and the product is not even a line item in a financial statement any more. Forget that in 1998 the company had north of 275,000 unique customers–more than half of IBM’s total 500,000 customer base at the time. I seriously doubt that there are even 200,000 AS/400 sites these days, and my guess is that even if there were, only 20 percent of them are buying and investing in the platform in anything resembling an aggressive manner. (Just like the IBM System z mainframe, by the way.) And for all IBM likes to talk about its Unix prowess, the company only had around 100,000 RS/6000 customers a decade ago–and that included a lot of workstation and modest server shops, not back-office stuff like what AS/400s did. I find myself wondering if IBM started with nearly 400,000 midrange customers and by its short-sightedness in the way it has treated its loyal AS/400 shops, maybe cut that base by a third. Or maybe in half. Sure, the revenue stream is still around $5 billion a year, but guess who got those other customers?
I can tell you that the IBMer in charge of the AS/400 division never took Microsoft’s Windows NT seriously at the time when it was ramping up and getting better in the 1990s, but by the early 2000s, it became pretty clear that Windows was taking a big bite out of the midrange business. And continues to gain ground. So an IBM obsessed about Unix and basically ignoring Linux, left its rear guard exposed and Windows attacked.
I happen to think this was a stupid way to play the midrange game. I said so then, I have said it repeatedly for the past 15 years, and I am saying it again today.
Back to the differences between the AS/400 and other proprietary platforms. Keeping the AS/400 platform alive and moving ahead on each successive generation of Power-based servers is a lot better treatment than Tru64, HP-UX, OpenVMS, and NonStop customers from Digital, Compaq, and then HP as each took control of the platforms. Unisys has done a remarkably good job of modernizing its mainframes and keeping them relevant in an X64 world, but ran out of money to do a proper and final transition. Don’t get me started on Sun, because, quite frankly, no one–and I mean no one–knows how the Sparc/Solaris platform is going to turn out. I don’t think Sun or Oracle have a clue yet, and Fujitsu is not helping with its increasing emphasis on X64 iron and lack of enthusiasm for its Sparc64 chips, which Sun and Fujitsu both use in their combined midrange and high-end server lines.
Here are some initial thoughts I have about a resurgence of the AS/400. Not in any particular order, except that this is how they are coming out of my head.
1. It’s the AS/400, and that actually means something to a lot of people. We still call this newsletter The Four Hundred, as we have for more than 20 years now. And when we are talking, we still call the box the AS/400 or the ‘400 for short. People only change their names when they get married or they are guilty of something. I can understand that the RS/6000 and AS/400s are on a common hardware platform, but they have been on one since 1997, in fact. This didn’t happen in 2008. Convergence helped one brand and hurt another. I would argue that such short-sightedness actually hurt IBM. And that this simple act–of calling the machine when configured with OS/400 by its proper name–would go a long way toward showing the AS/400 faithful that IBM, in fact, cares and listens. (I had a lot rougher language in that prior sentence, but this is a family newsletter.)
Hewlett-Packard gets this branding issue. Do you realize that the Compaq brand is still alive on consumer PCs because people recognize and like it? Or that 10 years from now, I will be talking about the ProLiant nanocube qubit quantum server, the 21st generation of ProLiant servers? I don’t know why IBM pays marketing people so much money to build a brand and then spends so much energy trying to destroy it.
But please do all of us out here in AS/400 Land a favor and show a little respect for us and for your own product. Change the name back to the AS/400. And the definite article is not optional. That was freaking idiotic, too. We’re not British, and we’re not in hospital. I love you Malcolm, but really. Why mess with something that works?
2. Get people onto Power6 machines and i 6.1 yesterday. This is a big one. I realize that IBM no doubt had technical and good reasons to radically alter the licensed internal code underlying OS/400 with i 6.1. The last time this happened was the jump from CISC to RISC processors in 1995, and only other time was the jump from the System/38 to the AS/400 in 1988. But here’s the problem, and one the AIX people didn’t get. By doing this now, at the same time as the Power Systems convergence, the Power Systems division broke the spoken and promised covenant with AS/400 shops the world over that they would never have to recompile their code to work on new iron.
IBM has kept that promise through very clever compiler technology over the past two decades. But IBM broke the covenant, not the customers, and that means Big Blue has to cushion the blow. In 1995, the PowerPC machines offered twice the bang for the buck as the prior CISC machines, and that was enough to cover the hassle of dealing with recompilations. By the way, the AS/400 revenues exploded from 1996 through 1998. And it wasn’t a coincidence. And when the 5250 software tax was instigated about then, sales fell every year for a decade. It doesn’t take Einstein to figure this out.
IBM needs to move the entire base to i 6.1 and Power6 or Power7 machinery, all at once. Give away the servers, give away migration services. Whatever. Get people current and excited.
3. Integration does not mean bundling. Having an integrated system is wonderful. It is what all system makers aspire to. Oracle understands this, as it chases Sun. Cisco Systems understands this, and it is a newbie to the server racket. How on earth did IBM’s AS/400 division forget this? Oh, right. There is no AS/400 division to explain to the AIX folks who took over everything that there was–and still is–a completely different way of selling systems and supporting applications that doesn’t involve handing over control of database and middleware software to Oracle and application software to Oracle and SAP. The Power Systems division has a Unix bias that is severely hurting the AS/400 business.
Yes, of course IBM needs to keep integrating new technologies with the AS/400 and its OS/400 operating system, but it needs to keep them unbundled and optional. The database itself should be optional and not included with the OS, for instance. On all machines and all processors and logical partitions inside a machine. Break out each item of the software stack–every feature of the OS, plus the database, compilers, tools, and so on–and give it a price on each machine.
4. Cut prices and compete, make it up with services. I explained this so many times you are probably tired of hearing it. But IBM could charge next to nothing or give the iron away and make up the rest in software licensing and recurring services fees. The amount of money customers spend over the course of three, four, or five years would not change a bit, but rather than have to fight to upgrade every three years and then live without budget for the next three years, the costs would be spread out monthly. (I will delve into this in more detail in a future issue of this newsletter.)
As I have said many times, the AS/400 is the underdog in the midrange space right now, and the company has to start behaving that way. This means aggressive investment in new technologies and cut-throat pricing. This worked brilliantly for the AS/400 in the late 1980s and early 1990s. It can work again in the 2010s with a whole new group of tens of millions of small and medium businesses who don’t care what their applications are programmed in so long as they run the factory, the warehouse, and keep the books–provided the price is right, of course. And that price? Damned near free.
5. Legacy is not a dirty word. Ask yourself this: Which platform of IBM’s is the second-easiest one to dump after its System x X64-based servers? It isn’t a Power Systems i box, ladies and gents, and it ain’t a mainframe, either. Which one has probably, over the past 20 years, contributed less to the bottom line? (Aside from those commodity X86 and X64 boxes, of course, which are useless in terms of profits.) Uh-huh. Thought so.
I am proud of my legacy, following the midrange for two decades. And IBM should be proud of the legacy of the AS/400, as well as its future. RPG and DB2/400 and OS/400 are not dirty words. Use them. Tell people it is OK to use them. Explain why they are better than Windows or Unix or Linux, and if they are not, well, make them so. OS/400 was arguably a lot better than the competition when it came out in 1988, and it took over the midrange.
I am just getting started. But right now I have to get out of here on a Friday night and take the wife to Town Hall to see Garrison Keillor’s A Prairie Home Companion, which I listen to a lot of Sundays while I am cooking but I have never seen live. Maybe a good place for the Power Systems people to start is to get out of the rarified air of Somers and Armonk, New York, and get some Minnesota entertainment and storytelling. It might be a little old-fashioned, but it is solid.