RealTime IT News

Jim Stallings, System z General Manager, IBM

Jim Stallings Just a few months ago, Jim Stallings led the open standards and IP income strategy for IBM's technology and intellectual property division.

But then IBM triggered an executive shuffle and relocated Stallings, who has a track record of providing a boost wherever he is placed, to head the systems group's System z mainframe group.

The change seems to have suited Stallings well. The charismatic executive likes to meet with customers to extol the virtues of mainframes and vowed to meet 100 customers in 100 days.

After a recent event celebrating IBM's mainframe momentum and new software tools, Stallings discussed the strategy with

Q: IBM's server competitors and even IBM itself are pushing modern, modular servers, such as blades. How do you know where and when to sell the mainframe and when to sell the blade?

The answer lies with the customer. The answer is: both. It's not going to be one or the other. This isn't about who wins. Customers are growing, and what they're finding is, there are certain workloads that, because of price, cost, licensing, ease of use, they are better served on an edge of network server like a blade.

But there is also a set of applications that run better on the mainframe. The guys that lose are the ones in the middle. They're basically trying to stretch a Unix server to do, from a cost standpoint, what a blade can do and it can't, and from an availability standpoint can do, and it can't.

So that's what's getting cleaved off. Sun Solaris and HP-UX are getting cleaved off on these edges where they're trying to stretch into these environments. Anyone who goes running around and says either mainframes or blades are going away, is trying to substantiate something that's in the middle.

We realize that there are certain applications where it makes sense to scale them out for customers. But we're not saying that at the end of the day you get rid of the mainframe and you can spread it all out. Physically, you can't. You can't take 900,000 users simultaneously and run them in one physical location with Intel blades. I don't care how big the server farm is.

The mainframe doesn't serve all environments. It's cost-prohibitive in some environments, but not necessarily in all startups despite what people say.

I just got a note this morning from a startup that bought a z890 mainframe. At $100,000, they can afford it.

I've started up a company from scratch and spent a million dollars on IT the first day, so $100,000 for a mainframe is pretty good. The company is a Web-based company, so they're thinking scale from day one.

Q: I wouldn't think a startup has the staff to maintain a mainframe. How many people does it take to maintain that?

One person.

Q: One person must seem like a dream compared to the care needed for mainframes of years past. Are mainframes getting easier to maintain?

More self managing. You go to CitiGroup and there are a lot of mainframe engineers and support people, but it's small compared to the shops with Unix servers that have literally hundreds of people to run them.

It is the No. 1 conversation that I have with customers when I first meet them. They say: "How can you take humans out of this?" They want mainframes autonomic, self managing, self-healing. The mainframe tracks, captures and stores literally tens of thousands of messages about what's going on in the system. Some of them are error messages.

There's this aggregation of data that can point to what's normal, what a normal day looks like, and then there's a zone of behavior in between.

We believe that given all of the thousands of messages that get thrown about the health and state of the mainframe and the applications and the I/O, that we can create this image of system health that is invisible and has very low overhead, but gives them some predictive indicator that something is about to happen.

This is something that's beyond humans to be able to predict. There is so much data, and what human can analyze it fast enough to do something with it?

Q: You mentioned earlier today that some companies, such as Google, told you they eschew metrics like speeds and feeds for newer equations like kilowatts per hour. Is Google using IBM mainframes or were you making a sales pitch?

No, they don't use mainframes, and I called them up and asked them why. I told them at some point they're going to have to hire the equivalent of India to run all the Intel boxes they have. That's when they said they measure computing costs by kilowatt-per-hour.

We're going into new markets and industries. Search is a big market and industry, so a lot of the players are scale-out companies. I think they all opted for cheap servers versus expensive when they began, but now they're getting into scale issues and human capital issues.

They've got more people in the IT department than the marketing department. So the marketing guy says: "You're allocating a cost that's bigger than my staff. What's up?" What wasn't on the chart is the energy cost.

Part of what's invisible now is the energy costs because CIOs are packaging it in. If you called it out as energy, it's a big red slice. Compare that in a TCO sketch against the mainframe.

Q: How do you get a newfangled company like Google to buy your mainframes at a time when rivals like Sun and HP are constantly saying the mainframe is dead?

The numbers say the mainframe is not dead. We've taken share and made Scott McNealy quit. Carly Fiorina quit. They got run out of town. These people lose their jobs. They don't lose their job when things are going well. They lose their jobs when they lose in the market place.

This isn't just stock price, it's market share. The mainframe has taken out a lot of Sun servers and a lot of Intel-based Compaq machines over time. There was a period of time in the mid-1990s when we were losing to these companies. Then the Internet bubble imploded because people said "this is about revenue numbers and not profit." And it all became profit and everybody left town. Sun never recovered because it's about value now, not speeds and feeds and clock rates.

Customers don't care. What we're swimming against is the CIOs in a lot of the big enterprises today that come from a generation where they don't necessarily have mainframe skill sets.

We're going to go very aggressive on this over the next several quarters, discussing how the value of the mainframe applies to new industries, new markets and new customers.

Quite frankly, and part of this is our fault, people just assume that mainframes are only for the people that have them. That existing customers are the only people that can afford them and that they're the only people that have the skills to maintain them.

We're telling our current customers that if they go and migrate from the mainframe to Solaris, they're going to pay to retrain and retool and their costs are going to go up. We're going to be in the market hard on this over the next few months.