From IBM server architecture to the Xbox

Legend is an overused word these days, but if 'industry legend' is a phrase that ever applied to someone who worked in the computer business then it would be applied to Frank Soltis – the man who has guided IBM's architecture strategy for more than 20 years.

Soltis rose to fame when he saved IBM's mid-range strategy with the development of the AS400 "Vax killer" in the 1980s and continued when he acted as rescuer again in the 1990s, when he found a lasting place for IBM's Power PC mid-range.

Since then his genius for computer architecture has spread beyond IBM and across the computer business.

If you use an Apple Mac, that is based on a Soltis architecture. Nintendo? The Sony PlayStation? They're Soltis again. The Xbox? "The Xbox 360 is due out about now," says Soltis, in an interview with ZDNet UK this week, and then smiles. "I really like that name, the 360."

Soltis' amusement at the irony of the latest Xbox sharing its number with the mainframe that established modern computing, the IBM 360, is understandable.

But for many, the idea that Frank Soltis is intimately involved with the games business may be hard to grasp. After all, this is the man who built the ultimate workhorse system, the AS400, which blossomed into the IBM iSeries. Like the AS400 before it, the iSeries is the machine that runs thousands of car dealerships up and down the country, that does the really boring invoice processing in travel agents around the globe, that churns away, anonymously, in millions of different businesses. It is the machine of which Soltis is most proud.

Soltis comes from Rochester, Minnesota, and so does the iSeries. That middle-America pedigree is entirely the image that Soltis sought and now cherishes.

For Soltis, there is nothing wrong with "dull". Being dull is a virtue that he has been able to exploit to such a degree that IBM's entire corporate IT architectural strategy is now based in that small city. Mainframes, mid-range, PCs — and as Soltis says, "we may have sold the PC business but we are still in the PC business" — are now all thought out and developed there by the "4,500, maybe 5,000" IBM employees in town.

It's called "Fortress Rochester" by the people who work there and the IBM people who don't, and Soltis thinks the name is absolutely right. From the Fortress he has fought many battles against others, who hail from more fashionable areas — Silicon Valley, New York, Texas — and kept the Fortress true to itself until it has become the dominant force in the largest IT company in the world.

The story of the establishment of the Fortress is so good it is worth the retelling. In the late 1980s, IBM had a problem. The new kid on the block, the Digital Equipment Company (DEC) — which would be taken over by Compaq, which would in turn fall to HP — was eating into IBM's markets for minicomputers with its VAX architecture. IBM, in big...

...company style, put out a tender to all its own design centres to come up with something to beat DEC. All the groups contribute flip-charts and flow-charts and descriptions of how the system would work.

Soltis walked into the meeting at IBM HQ with a working prototype.

He plugged it in and showed them the system in action. His group won the contract and the rest, as they say, is history. The AS400 was put together in two years and within a year after its launch it was the best-selling system in the history of IBM.

It is such a good story it is often thought apocryphal, so is it true? "That's about how it happened," admits Soltis. "Although, I never saw it as a VAX killer."

The Rochester plant grew and took on more responsibilities. And grew some more. Two years ago, IBM put all the development in the hands of Rochester. And in all that time, Soltis showed no sign of wanting to move anywhere else.

It is said that Soltis believed that Rochester was a great place to do development "because there isn't too much else to do and the winter cold makes people want to work". But talking to him, it becomes obvious that there is something much more. There is a streak in Soltis that says that while most Americans in the IT business would much rather live in Florida or California, Washington State or Massachusetts — anywhere but Minnesota — the place to craft solid, hard-working, reliable systems and architectures is in a place that lives those virtues.

Soltis does not like to leave Minnesota. "I really don't like travelling," he told ZDNet UK on one of his rare trips away from home. "I am happiest back home with the other engineers. Just put me in the lab and leave me."

But this diffidence is something he now finds he has to learn to deal with. The success of Rochester has taken Soltis from the shadows, and put him front and centre. Having fought to get his group recognised, the situation has now swung completely on the opposite direction to the extent that his division now carries the hopes of all of IBM.

That means more air miles for Soltis but he does not seem the type to grouse about that too much. Not when he has the chance to talk to more people about his favourite subject — computer architecture.

"The big breakthrough for us was the development of the Power architecture," he explains. "It was the first fully virtualised architecture. You know, people talk about that a lot, but we did it. It means the software is completely separated from the hardware. We saw that early on. The hardware changes all the time, but the software is hard, and expensive to change."

When Soltis was called in by IBM's senior management five years ago he was tasked with developing a five-year plan for IBM, he told them he would extend that virtual architecture across all of IBM's systems. The result is that IBM is now in a position where it can build solutions much more quickly and easily than it could before.

"With our virtualised architecture you can swap anything around. You can use the same components [in a PC or a mainframe] and...

... we aim for the maximum flexibility. You can change something, the software will still run."

Soltis points out the virtues of that way of working and the problems that can occur. "I really give credit to Intel for developing an architecture in the 70s that still works. Xeon is fine. Then they developed Itanium with HP and they put in a RISC processor and that is great but then you look at the trade-offs. You have to change your software to work with Itanium and that is the trade-off. Changing the software is expensive."

Soltis' design goals are all around avoiding trade-offs. "The iSeries goes from small to very large and you don't have to change anything." Virtualisation is one of Soltis' key themes but always within the context of efficiency.

You soon get the impression that this is a man who hates inefficiency and the subject moves on to processor design, which Soltis freely admits is not his field. "But you look at processors and everything is about [clock] speed," says Soltis. "But we know we are up against a speed limit. To get that speed the circuits are so small that they are only a few atoms across and are so small and they run so fast they can't carry a signal so what do you get — heat. Our processors aren't the fastest in megahertz, or gigahertz, but the systems run faster than you expect."

It's all about the right architecture, he says. And the trade-offs; and at this point, Soltis moves back to the Xbox. "A games system is all about performance, but in a different way from a business system. In a games system it is the front-end bus that is important. On a business system that can be as much as 800MBps. On the [Xbox 360] it is 31GBps. But the software to use that is not here yet and that takes time for people to catch-up. It is a trade-off."

Games machine or mainframe system, Soltis will talk with authority and enthusiasm on the architecture of either. And that goes for supercomputers as well. The Rochester realm under Soltis developed IBM's Blue Gene which is "exploring the frontiers in supercomputing, computer architecture and in the software required to program and control massively parallel systems" and "in the use of computation to advance our understanding of important biological processes such as protein folding". Soltis sees it as dealing with the same issues as in an Xbox or an iSeries machine, just with different trade-offs.

"Supercomputing is fascinating because the architecture is one where the performance is the most important thing and how much it costs is less important" he says.

Now Soltis is focused on the next five years at IBM. The architectures from desktop to supercomputer are all working in the way he wanted with maximum efficiency and minimum redundancy. IBM's customers are happy, at lease with the workhorse iSeries, which, according to Soltis "grew 25 percent last year, which is not bad".

Soltis admits that he finds his role at IBM a "great responsibility" but gives every appearance of enjoying what he is doing, if it wasn't for the travelling. After London, he must visit the troops in more foreign corners. Is he looking forward to getting back to Minnesota? "Oh I am," he says.