Server Ecosystems: Take a Ride on a Slide (Continued)

<<< Click to return to the first part of this story

But I want you to notice something important: while the Unix server market was collapsing from 2000 through 2004, the IBM mainframe market stabilized and has perhaps even rebounded a little (if you average it out). If IBM can stabilize the mainframe market by making the mainframe relevant to an even a much smaller base of customers (maybe 15,000 to 20,000 machines at maybe 10,000 sites), then it can certainly do this for the iSeries, which has a customer base that is at least an order of magnitude larger and an installed base of machinery that is more than 20 times as large. The number of customers matters as much as the machine count and the revenue count. And when it comes to customer count, about half of IBM's customers are using an iSeries. So don't get all freaked out by the AS/400 Slide:

But by some means, get a little freaked out. This has been a very tough business for IBM to maintain its hardware revenue stream in. So tough that, in fact, it has not been able to do it any better than it was able to do so with the mainframe. As best I can figure, the AS/400 Slide is about three years out of phase with the Mainframe Slide, with a slightly different shape to the curve. But take heart. When I get around to drawing the Unix Slide, it will be about 10 years out of phase with the AS/400 Slide, but the curve will not look very much different. And I think the Windows Slide for Windows-based hardware sales will peak soon and then begin a downward trend, perhaps within a few years. Linux will be growing for some time to come, and even Sun's Solaris Unix variant might see some traction now that Sun is supporting X86 iron. I think right now the two dominant server architectures are Power and X64, and more platforms and applications will move toward them, even as Sparc and Itanium platforms continue to be sold aggressively and profitably for the next couple of years.

The thing to remember is the OS/400 ecosystem, which is why I started out trying to gauge that first. IBM's AS/400-iSeries hardware sales (the slide includes systems/servers plus internal and external disk arrays) are down by a factor of three from the peak in 1991 at just under $5 billion, I don't think the OS/400 hardware, software, and services ecosystem has contracted by any more than a factor of two. And when you factor in the value of programmers, managers, system administrators, and other personnel at OS/400 shops, who are a key component--if not the key component--of the OS/400-based solutions that companies use to run their businesses, the ecosystem may not have even contracted that far. (It has certainly contracted, however, since there are certainly fewer OS/400 experts and fewer OS/400 sites, even though sites tend to pay more for employees than they did a decade ago.)

For my final thought on this whole ecosystem issue, here is something very important to contemplate: What is the economic value, in terms of the OS/400 ecosystem, of the things you do not have to pay for? Do you know how many database administrators there are in the Unix world? Zillions. How many at OS/400 shops? Few. OS/400 was designed from the ground up to provide relational database capabilities, a relatively simple set of business logic programming languages (RPG and COBOL), and an integrated system that required very few people to manage.

If IBM wants to boost the iSeries business, it needs to make these things--not Java, not Linux, not Windows on an xSeries co-processor--not just acceptable, but cool again. Why isn't RPG cool? It is a perfectly reasonable business programming language that should be used by businesses instead of Java because Java requires rocket scientists. If I were running IBM and I wanted to pump up the iSeries, I would do a few things. First, I would take the RPG language open source and demonstrate that it is the best language for building business applications. Show this. Prove it. You all believe it is. Demonstrate this.

I would then foster a community of open source RPG developers who would create modules that could be connected together to create applications. Then, I would create a new line of entry iSeries servers that cost $2,500 and $5,000--perhaps running in emulation mode on X64 servers, perhaps running in emulation mode on entry Power servers that do not currently bear the iSeries label. Real servers, not crippled machines. Then I would give away commercial licenses of OS/400 away for free and charge for tech support, using the Linux and now OpenSolaris models. Just let people have the code and they will pay for support, and you can make it up in volume.

The key is to get that ecosystem growing, and you can only do that by making the installed base bigger or getting customers to buy bigger systems. IBM has engineered gargantuan iSeries machines, but very few customers need them. So aggregate sales have fallen along the price/performance curve just as fewer customers need to buy a machine every two or three years. Small businesses do not buy big boxes to do their books, and IBM has not faced this fact in either the pSeries or the iSeries line. This has not had much of an effect on the pSeries, but it has been disastrous for the iSeries. And converging the pSeries into the iSeries, which really began in 1997 and culminated with the i5/p5 line last year, into the zSeries line, which I still think will happen sometime in the Power6 generation, is not going to fix the iSeries. By all means do that, IBM, if it makes sense for the mainframe.

All I know is that a static installed base leads to a declining ecosystem in a world where other bases are growing; a declining installed base leads to a rapidly declining ecosystem; and a growing installed base leads to a growing ecosystem. Unfortunately, it takes an exploding installed base--like Linux now, Windows a few years ago, and Unix a decade ago--to create a rapidly growing ecosystem. It is really that simple. Adding AIX and Linux on the iSeries and Windows on a co-processor is not enough. Look at the numbers.

So, IBM, you have to get that iSeries installed base growing again, and growing fast. While the i5 line as currently configured is a great improvement over the iSeries from 2003 and a huge improvement over machines from earlier years, it is not like the move from CISC to RISC in 1995. We need something radical, something explosive, and that is a lot to ask for in a dubious economy from an IBM that is distracted by troubles in the services business and content with its overall portfolio of servers. But, on behalf of all of us in the OS/400 ecosystem, I am asking just that.

IBM, get a little crazy, get a lot more creative, and get with the program. Make some noise. Say these words to yourself over and over until you believe it: OS/400 is better than Linux, and RPG is more appropriate for ERP applications than Java. 5250 is an efficient, secure protocol on which to build Web applications, and what we really need is to make Web-style programming easier. Say it loud and say it proud. A quarter million midrange businesses are not wrong, but your approach to building, delivering, and pricing midrange computers is out of phase with this century and with millions of other companies. Start an RPG and OS/400 development community, involving--deeply involving--the customers; make the software open source under a license that makes sense--it doesn't have to be GPL. Do something completely nuts, like make an RPG virtual machine, if this makes technical sense. Distribute it for free to people who want to run RPG applications using an open source RPG compiler and, yes, open source RPG application programs. By all means make the iSeries Power server the best machine on which to run OS/400 and RPG with full-on, IBM-class, worldwide tech support, but make OS/400 and RPG available on inexpensive X64 iron. Go for volume, or no one is going to hear you.

Don't wait. Do something now. Let's slide up for a change.

RELATED STORIES

How Big Is the OS/400 Ecosystem?

The OS/400 Ecosystem, Part 2

How the Server Ecosystems Stack Up

TFH Flashback: Critical Mass, November 1993 issue