Frank Soltis Talks About Shared Technology

by Dan Burger

Soltis is also a frequent speaker at conferences and user group meetings. During a recent swing through the western United States, he was a guest speaker at back-to-back user group meetings in Orange County and San Diego, California. Soltis spoke with The Four Hundred between stops to talk about future plans for the iSeries and what's on the minds of iSeries users.

Where does the iSeries fit into IBM's plans as the eServer brand blurs the distinctions among the various server groups?

Within IBM we have made a conscious effort to share technology across all platforms. We started that about five years ago. IBM never used to share technology among server groups. It was a competitive environment by design, and [that was] thought to be a better way to foster technological advancements.

But long before there was an official eServer policy, there was a desire to converge systems--two computer systems into one, or three into one--but it was always a dismal failure. The reason was because it was not converging on the same technology. If you really want to do this, I told them, you're going to have to start with hardware and move up through the software. I was able to convince some folks at IBM this was a good idea.

My role was to begin the effort that became the 64-bit PowerPC. My view was that hardware technology would lead to common software. At the time this was formulating, we [the AS/400 group] were moving to the RISC processor and we had to rewrite the internals of the operating system. Every operating system in those days was written in a different proprietary language. There was no sharing.

We moved into C++, and then to Java, as a development language. So with the PowerPC and the same development language, we could have shared operating system code and shared middleware, and eventually shared applications.

We delivered the first RISC processor in the AS/400 in 1995, and the rest of the development groups said, "We're not interested in common technology."

That changed in 1996. The decision was made that we had to start sharing technology. The first step was to take the AS/400, load AIX on it, and rebrand it as the RS/6000. It began to ship in 1997. The hardware of the two platforms is almost identical, and has been since 1997.

Looking ahead, the next step is to move the mainframe to that hardware. That should happen within two and a half years. Internally we refer to the next generation mainframe as the Power 5. It will have the iSeries, the pSeries, and the zSeries all on the same hardware. This is the continuation of shared technology.

What other notable examples of shared technology would you say have advanced the iSeries?

With V5R1 we introduced a new hypervisor in the machine. A hypervisor really takes control of the logical partitioning, as well as the capability to run multiple operating systems. That's what allowed us to run Linux in a partition. That particular hardware and software technology results from hardware hooks that make that work in the PowerPC processor. We put that exact hypervisor code in the pSeries, last October. That same code will go into the mainframe in about two and a half years.

Doesn't that open the door for running Unix and Windows in a partition on the iSeries?

Potentially, that is the view that I've had of the platform over the years. I like to call the iSeries a "universal server." I would like to see it run applications from any operating system. And if that means running the operating system, which in most cases it does, so be it. Obviously, when you start crossing into operating systems that are owned by other companies, like Windows, you have to have their agreement.

I would love to see Windows running natively on PowerPC. I've gone out on a limb and predicted it will happen. There is interest on both sides. But this is a business decision between IBM and Microsoft.

Keep in mind, there are a lot of things I would like to see happen that are not necessarily in IBM's plans. A lot of these things, from a corporate standpoint, won't even require a decision for a couple of years.

Personally, I think it makes sense for both companies. Much of it has to do with scalability. From a Microsoft standpoint, PowerPC as a platform was really designed to be a server processor. It will out-scale what Intel is doing. Intel is really building a single-user workstation processor, even though they sell it in servers. And Microsoft understands that. So, if you really wanted a big Windows server, you need PowerPC.

From an IBM perspective, we would have the capability of the management components. I'm talking more than the iSeries, because you will be able to run in partitions on the pSeries or zSeries as the hardware and underpinnings start converging. So it would be feasible to run multiple copies of Windows in partitions, under the management characteristics of something like OS400 or another major operating system. We could show a very good business case for companies that are struggling to run a lot of servers.

We are doing it today with Linux, and we have customers who are using the integrated xSeries adapters to run Windows 2000 applications, which run their business. They have no AS/400 applications, but they buy iSeries to manage the Windows environment. And we have some that are doing the same thing with Linux. It's not a huge number, but it's one of those things that's beginning. More and more people are discovering that it works quite well.

And what about Unix running in a partition on the iSeries?

Keep in mind that the RS/6000 is literally an AS/400 with AIX loaded on top of it. When we did PASE on the iSeries, we seriously considered an implementation that would run full AIX. In those days we were using coprocessors. We didn't have full partitioning. So we decided that it was not the way to go for Unix. We created PASE, which is a runtime environment. It's not in a partition; it is part of OS/400.

When the hooks are put into AIX so that it can do full partitioning, maybe in one of the next several releases of AIX, it could become a business decision. We could do it today, from a technical standpoint. AIX would run under OS/400; it doesn't need all the pieces. This would not allow AIX to run in the primary partition, though, because not all those pieces are there yet.

I believe there is a business case for it, in terms of AIX as a specific Unix. Let's talk about a corporation that has an Oracle standard for database. The iSeries, of course, can't run Oracle. Well, Oracle runs on AIX, so that is an alternative.

It's not going to be a real surprise to me if we see something like that come along. I know there are business partners that like it. Some don't believe Linux is robust enough at the moment to take on those applications; AIX is. Technically, it is a no-brainer.

I also understand that Oracle is going to be available on Linux in the not too distant future.

Will the iSeries become a more job-specific server in the future?

Most of the things we are doing make it more general-purpose. One thing is the capability to run AIX. Some companies like to run Unix applications. That broadens the base for the iSeries.

Unfortunately, this news often gives a mixed message about what is happening to the AS/400. It's like when we brought out Java. In the early going, all you heard about was Java, Java, Java. The fallout from that was that some said, "Oh, my gosh, RPG is dead." That was not the intended message. The reality is that RPG is the development environment for iSeries; it has been and probably will continue to be. The same has been true with Linux. Some people fear that all applications will move to Linux. No! It's just another option.

It's hard to get the message across that we have a broad selection and we're making it broader.

If the base becomes broader, what does that do to RPG as the primary development language?

People concerned about application development ask me, "What should we do?"

My answer is, "Who are you writing this application for? If this is your company's application, and you develop in RPG, why would you want to write it in Java? If you want an application that runs across lots of platforms, then look at Java."

Most of it depends on what skills are in the shop. If it's RPG skills, then fine. But does this RPG application have to run on a Unix box? If it does, then it probably doesn't belong on RPG.

People assume everything is black and white. If you talk about Java, then RPG must be dead. I like to point out how many times COBOL has died over the years.