Certain industry executives who have been around a while never seem to get ruffled by the highs and lows of the industry. Economic downturn? Seen it before. Hot technology trend? Well, there's really not that much new here. Steve Mills, head of IBM's Systems and Software Group, is that kind of no-nonsense guy -- one who is seldom shy about giving his unvarnished opinion.
In this interview, conducted at last week's IBM Impact 2011 show in Las Vegas, Mills offered a candid assessment of HP's purported move to the cloud. Along the way, he resisted my attempts to suggest that cloud, mobile, and big data were conspiring to drive unprecedented change -- and offered one of the more down-to-earth explanations of cloud computing I've heard.
Last summer, IBM merged its Software Group -- which Mills already led -- with its Systems and Technology Group. As senior vice president and group executive of Systems and Software, Mills now has responsibility for all IBM products, making him one of the most powerful people at the company. We began the interview by talking about his new position.
Eric Knorr: You now have essentially double the responsibility you did before.
Steve Mills: I was spending a lot of time on hardware and hardware-related issues before, so it's not a new thing. The software team has had to spend a lot of time with hardware guys on hardware design -- and I'm spending more time on more aspects of the hardware design and pushing the hardware teams to pay more attention to workload-based characteristics.
Knorr: So that means you're involved in product development on both sides?
Mills: Right. All of product development, manufacturing, marketing, sales; it's over 100,000 people in lots of locations all over the world.
Knorr: Well, you're in a good position to answer this question. Recently, I've been hearing statements from executives around the industry that over next 5 years we're going to see more change than we saw during the last 25. Or words to that effect. Would you take it that far?
Mills: Well, there's no scorekeeper, you know? Change is always a matter of your own perspective in terms of where you are. Clearly, technology builds on technology, so there is an expansion phenomenon that occurs around technology. We all stand on the shoulders of giants.
So yes, the next 5 years will be profound in terms of new things that we see. And yet if you think about today -- pick 2010 and think of 2005 versus 2010 -- a lot changed, but a lot stayed the same. You saw things evolve more than absolute change. Handheld devices were certainly popular in 2005. People were running around with tablets back in the late '90s. They didn't perform that well and the model needed more work to get perfected. Now the whole market is abuzz with the tablet idea. It's not a new idea; it's just a bunch of things had to come together -- size, weight, the power, battery life, application functionality, things of this nature. Obviously the Web and high-bandwidth connectivity, all of which are required to make the iPad, or whatever your tablet choice is, work effectively.
Knorr: So you see the PC paradigm breaking up?
Mills: I think it has. I think that it's been breaking up for a long time. Those of us who are old enough remember that we lived in our Windows desktop. Our world was defined by the device and what you ran on the device and whatever server you were connected to -- or could get connected to, which weren't many back then. Your connectivity options were limited.
The Web changed all of that when Netscape exploded on the market. And since then we've watched bandwidth go up dramatically -- now rich media, TV, movies. Obviously mobility and high-bandwidth wireless have made tremendous additions. And we know high-bandwidth wireless is only going to go up.
I'm always focused on this issue of bandwidth and speed, because I know that people want to do more things with computing, with devices, with whatever, but bandwidth and speed have been a limiter. You saw IBM play "Jeopardy." So the capabilities that we built into that system, the Watson system, were built on top of things we've worked on at IBM literally for decades. The computer science structures that were used to create a learning environment, they were decades old. But we're at a point now where you could actually build something that's a reasonable size, you could stick it in a room. You couldn't have put all that together and gotten it to perform effectively a few years back. It's performance and bandwidth.
Your ability to store and manage ever larger amounts of data keeps going up and up. And so as these things widen out, new opportunities emerge that would have been hard -- not impossible to do before, but certainly hard to do economically. And I think that's really where most of the magic happens. It's not so much that invention has increased at some phenomenal pace, but innovation caused by the fact that the technology is cheaper, more accessible, and can perform faster widens the aperture of the new uses, new capabilities.
Knorr: Would you include in that the ability to handle huge amounts of unstructured data?
Mills: Yeah. That's the Watson model and certainly other big data projects. We're seeing an ever-growing interest in big data projects in all kinds of industries. That's going to increase pretty dramatically over the course of the next five years.
Knorr: There was a big emphasis on change at the [IBM Impact 2011] keynote this morning. What are some of the key decisions for CTOs right now? With the acceleration of change, the stakes immediately go up on the technology decisions you make. On the data center side, on the application infrastructure side, what are some of the key decisions now that people should be looking at?
Mills: There are two very fundamental, very interesting trends that are taking place. One is that the server side of the computing model is condensing. If I'm a CIO or CTO of a company, hopefully I know enough about the technology to recognize that I no longer need "n" number of data centers. I could turn that "n" perhaps into 2. I need a primary and a secondary, right? Sort of two primaries backing each other up. I can run all the IT for my entire company in two facilities and I can compress that footprint down -- consolidate work, virtualize my systems -- and that's pretty powerful in terms of improved economics.
A lot of companies recognize that and that's the path they're on. But some still have other ideas for various reasons. Sometimes they're caught up in false economics. But I'd say the general trend is toward server condensation, whereas the device trend is expansion. So you have this curious phenomenon: Servers become more compacted and dense and devices become ever more dispersed.
Knorr: So a big part of that consolidation on the hardware side is virtualization. If the "private cloud" means one indivisible thing, it means well-managed virtualization. Are CTOs thinking in terms of "cloud-enabling" everything? Or are they thinking more in terms of specific workloads?
Mills: They start with specific workloads. They start with development and test and then they move into other workloads, typically what you might characterize as non-mission critical.
Workloads that have a wide dispersion of how they get utilized or -- can I take certain production environments where I know I have a lot of vital capacity during key parts of the day and schedule work onto those systems? So cloud is really about dynamic provisioning and scheduling. Virtualization is an enabling technology that helps you deal with loading up your servers, but virtualization does not equal cloud and cloud doesn't equal virtualization.
Cloud is best defined as being dynamic provisioning and scheduling. In a perfect world you'd like it to be request-driven, meaning that you as the user, or you as a business, could get access to the facilities and services that you require when you need it and pay for it based upon that which you use.
Now, if you're a continuous user -- 24/7, always on, always consuming -- maybe these cloud ideas might not be best for you. Just because you're a constant user, you perhaps want more dedicated resources for your needs and requirements. If you're a high-frequency algorithmic trader, the idea that you have to share your trading activities with somebody else's activities doesn't make any sense to you. You say: "Wait a minute, I'm a millisecond-response-time guy. If you run other stuff on my machine you'll screw it up and I won't get the kind of response time I need."
So certain tasks and activities require dedicated resources, because they financially justify themselves. Some things are very compute-intensive, and you really use up the system, you don't have idle capacity. But with lots of stuff you do have idle capacity. So corporations want fungibility around their assets, they want to eliminate that idle capacity and have much more of a dynamic request-driven provisioning and scheduling infrastructure. And so virtualization is a step on the way, but it's merely an enabler. It's not core to the definition.
Now, what's the difference between private and public? Well, private is mine, the technologies are available to me, so I can deploy them and I can be just as dynamic as anybody is for my own in-house uses because that's how I choose to implement. Maybe I'm concerned about security, privacy. I don't think any of my outside suppliers can guarantee the quality of service I want -- they won't stand behind auditability, they won't help me certify Sarbanes-Oxley or other regulatory rules or requirements that I face. And so I'm not simply pushing everything out to the cloud just because that is the buzzword of the day. I'm not going to take all my work there.
But, by the way, if I'm a three-man startup, where else am I going to go [except the cloud]? But then again, I'm not worried about the Securities and Exchange Commission if I'm a three-man startup.
Knorr: The other thing is you're not going to start your own private cloud with a small business. It's a scale thing, right?
Mills: Yeah, it's a scale thing.
Knorr: How much of it would you say is really just about driving the cost out of operations?
Mills: I think that's the number one reason why this is appealing to CIOs. The interesting part about "cloud speak" is that many people want to isolate their discussion of cloud to only certain classes of companies. And my view is that's too narrow. Service bureaus emerged in the 1960s. I mean, ADP is one of the industry's biggest and most successful cloud companies.
Knorr: And earliest, right?
Mills: Yeah. If I'm in the accounting department and I use ADP, ADP is my cloud company. If I'm in the sales department maybe I'm using Salesforce.com. They're my cloud company. Now, Salesforce claims ownership to the term, ADP doesn't. Well, what's the difference between the two?
Knorr: I think we're talking about where we are now. If you're talking about making decisions that you're going to have to live with for many years, don't you have to start thinking about questions like: Which virtualization solution am I going to choose? The ultimate potential is when you start being able to cross boundaries with your workloads with the so-called hybrid cloud, where you may be able to start offloading those virtual machines to service providers.
Mills: Particularly if your business has a lot of variability to it. Retailers are the easiest example, where they need lots of capacity from September to December, but not nearly as much for the rest of the year. Why not overflow?
In fact, what we are out delivering are hybrid cloud services today. We have a whole set of offerings which are business process, but we also have generic processing cloud capability. We launched our Smart Business Cloud last week. We've got things like LotusLive, we've got software-as-a-service offerings and things like Sterling Commerce and Core Metrics and Initiate and a bunch of other things that we've either bought or added to the portfolio.
Everyone is talking about the cloud being new. Whenever I say ADP, it's like you throw a wet blanket on all of the tech industry double-speak. Oh, now you're taking all the fun out of it. The VCs hate the discussion of ADP and other online service providers, because the spin that they're doing on their cloud companies then sort of diminishes.
It's not really the next new thing, but you can do more of it today than ever before because bandwidth has gone up, performance is better, standards have improved. And businesses, I think -- even large businesses -- have become ever more open to this idea of a federated service model, where not all of the compute that I do needs to be on my premise and inside my four walls.
Knorr: With the cloud, do you think there's more opportunity for verticalization, especially with IBM, through its partners?