Steve Mills, senior vice president and group executive at IBM, runs the company's US$13.6 billion software business. At IBM's headquarters in Armonk, New York, Mills met with a group of Computerworld editors to discuss a software strategy that derives $11 billion of that $13.6 billion from middleware.
Q:Why is such a huge percentage of your software investment in middleware?
We do that because it's a good business - we have shareholders to satisfy. Middleware has always been important to customers - they need to do transactions, data management, development tools, systems management, security, collaboration.
We continue to invest in MVS on the mainframe and OS/400 for the iSeries platform. There are investments we have to make in support of the hardware platforms -- the operating systems are important. But the bulk of the money goes into middleware. And that's where the growth comes from.
Q: Is it a matter of using middleware to keep legacy systems alive as opposed to migrating your users off those systems?
Yes and no. Will you ever replace what's there? Eventually you will. For any business, what's the cost of rip-and-replace, and what will the benefits be? There's a need to sustain the legacy -- legacy is not a bad word in this context. It's a reality.
Q: Hewlett-Packard (HP) is migrating users off the proprietary e3000. At what point will you be in the same position of recommending to your iSeries (formerly AS/400) users that they migrate off?
Probably not in my lifetime. This is the "mainframe is a dinosaur" idea -- things old are not things good. No one has yet built an operating environment that's more reliable and more stable than the OS/400 iSeries system. And over the course of the last year, the installed capacity of that iSeries environment has been rising in significant double digits. That platform is a growth platform in the marketplace. Why would you move off it?
Q: You're talking about existing iSeries users. Are you getting any new users?
There are some new users -- there are new businesses that decide to deploy the iSeries. There are even businesses that get off of it and come back to it.
This is a very strong business for the IBM company -- it continues to grow. So the whole world is not going in one direction. This is part of the illusion in the IT industry, often promoted by vendors, not by customers: Vendors would like to declare that everyone's going in one direction. The market is far more diverse than that, and decisions that are made are made in a very different context than what the vendors would often have you believe.
The challenge that businesses have is not taking things away -- it's getting things to work together. A disproportionate percent of all IT spending goes to integrating the environments that are there. Eighty percent of what customers spend is related to that which they have today, not anything new they're going to buy.
Q: How would you characterize your operating system business in general, and AIX in particular?
The operating system business, which is a multi-billion-dollar business for us, in recent years has essentially been a flat business. AIX has been a flat business. We're supporting Linux on the pSeries today, and I think we'll see more and more customers over time decide to run the pSeries with Linux.
But no one's ever made any money in Unix operating systems. Unix has always been an enabler of the hardware platform. We went into the RISC hardware business, and we needed an operating system. Arguably we could have literally written one from scratch. We took the Unix System V kernel and we made modifications, the same as everybody else did in the eighties.
In the current decade, the market clearly is moving down the path toward Linux. Linux will run on a variety of different chip sets; it certainly runs quite nicely on the Power chips. In the future, we'll see more and more customers run Linux on Power instead of running AIX on Power, and to us that's fine. AIX has always been an enabler. All the money in the Unix market has been made in hardware, in middleware on top of Unix, in applications and services on top of Unix. But not in Unix operating systems -- that's not where the payback has been.
Q: So is it a matter of shifting your investment from AIX to Linux?
We're investing in both AIX and Linux today. We'll continue to invest in AIX probably for as long as I'm on the planet. More and more people, we believe, will choose Linux to run on Power vs. AIX on Power over time as Linux continues to increase in maturity and more and more things run on Linux on Power. So that's a logical progression of usage, and that's fine with us. We'll reach an inflection point where it'll be appropriate to spend less on AIX to enable the hardware.
Q: How much of your attention is the SCO Group lawsuit against IBM getting?
Obviously, it gets attention from a variety of people in IBM. We have to deal with that as one of these things that you just have to deal with in business. But the fact that SCO has got this thing going isn't stopping our activity in the marketplace. We haven't in any way moved away from Linux as a major investment; we haven't moved away from selling and promoting Linux.
Q: Have you personally spoken to anyone at SCO to try to work it out?
I have not.
Q: Do you foresee that happening?
They sued us ahead of them coming to us. They did not declare they had an issue: "IBM, would you work with us to resolve this issue?" They just simply filed suit. They decided to take the initiative away from the business people and hand it over to the lawyers. If you want to sue me, you have to deal with my lawyers. If you want to talk about a business problem, we can work on a business problem. If you have a proposal based upon some set of facts where you think IBM has done something and you'd like IBM to correct that, you could have come forward to us and showed us what the issue was and asked us to correct it.
We haven't seen any code that we have taken from SCO. You have to make the pilgrimage to Utah and see this stuff under armed guard or something -- I'm not really sure how they do these inspections, but we're not involved in that. It'll all come down to what goes to court.
Q: Your "on-demand" initiative appears to many users to be the same thing as the adaptive and utility computing models being extolled by such companies as Sun Microsystems , HP and Oracle. In one way or another aren't you all talking about similar kinds of things?
I don't think we are -- absolutely not. In October of last year we very clearly and definitively defined on-demand: A business whose processes are integrated from end to end -- it's about business process integration.
Then there are other things. There are issues of data center efficiency that customers want to talk to us about. But on-demand is not defined as data center efficiency. Others have data center efficiency initiatives that they're running -- that's what Sun is talking about; that's what HP is certainly talking about; it's somewhat hard to tell what Oracle is talking about. We're running a business process integration initiative.
Q: There was a recent report, posted by the Computer & Communications Industry Association (an anti-Microsoft organization), that said Microsoft's dominance on the desktop was a big security threat (see story). How are concerns about security affecting customer demand and customer choices on infrastructure and software?
This is something that's going to be fascinating to watch over the course of the next year or so. The accumulated effect of virus attacks and worms (has) brought about a lot of anxiety among businesses around the world on the porous nature of their environment and the cost associated with remediating these things when they occur. Microsoft, for reasons ... of their technology, which was never designed for this type of interconnected, public-facing Internet environment, combined with the fact that they are an obvious target, has raised all these concerns. Businesses are asking the question, What should we do about this? What kinds of remedies? There are no perfect answers. Many of the answers are procedural in nature, and some are technological.
For example, from a technical standpoint, if you have Unix systems facing the public Internet, you have the ability to dynamically change those environments on the fly. The symbolic referencing structures within the Unix environment make that environment more flexible in terms of being able to make changes to it, to block and prevent things coming in. Businesses that have Unix systems facing the public Internet are somewhat better off then those that have Windows server systems.
Q: You don't think it's dramatically different?
I think it is dramatically different in the context of manageability and the ability to, on an emergency basis, update all of your servers on the fly to prevent certain things from intruding into your environment. It is a big deal. We do it in IBM. We don't have Windows systems facing the public Internet; we have Unix systems.
Q: So is improving security just a matter of putting in better administrative controls, or is your safest bet to have a heterogeneous environment?
I don't think that the heterogeneity per se is a requirement. The systems-characteristics issues do play a role in this. I can be entirely Unix-based and have a more flexible capability to patch and fix and update and correct security flaws than if I have a Windows-based environment.The issue of heterogeneity, as depicted in what's been written recently, I don't believe is actually correct. The root-cause problems are the massive number of security flaws and holes in Windows, combined with the Windows design point, which is monolithic and inflexible, which leads to an inability to make on-the-fly corrections.
Q: Is IBM doing anything at all to offer alternatives to the Windows desktop?
Thin clients, yes -- we have a big portal initiative. We recommend to customers every day that they try to move the maximum number of users in their organization to a portal-based environment -- business workers, branch people, clerical people, plant-floor people. How many of your people need Office? The lowest number I've ever heard for thick client support is US$5,000 per user per year -- $5,000 to $12,000 (per year in support cost), that's kind of the spectrum. The big cost savings is not in getting off Microsoft Office and going to StarOffice; the big cost savings is, Can I get to a thin client? The cost of the client is tied up in the labor, not in the cost of the client box or the software that runs on it.
Q: Where do you see your portal going in the next year or two?
The WebSphere portal product has been growing 50% compounded and continues to do so. So there is great popularity around this notion of portalizing application interfaces in order to improve user productivity.
Q: Do you see an emergence of a new assault on the Microsoft desktop?
The browser-based approach, the portal-based approach, is the direction we see lots of businesses wanting to go in. There is anxiety over what they are spending on Office and how big will this desktop environment become.
- This interview was conducted by Computerworld's Patrick Thibodeau, Tommy Peterson, Don Tennant and Maryfran Johnson.