Ray Ozzie's view from the clouds
In an interview with CNET, Microsoft's software chief talks about Windows Azure, lessons learned from the Sidekick fiasco, and a future in which devices record everything imaginable.
LOS ANGELES--When Ray Ozzie penned his Internet Services Disruption memo back in 2005, he had a pretty good idea where the computing world was going. He just didn't know how Microsoft was going to get there.
While many are ready to write off Microsoft as an declining icon of computing's last generation, Ozzie sees Microsoft positioned to leapfrog some of the companies that tend to be thought of as the leaders of the cloud computing world--names like Amazon, Salesforce and Google.
"I will never, ever, utter the words 'mission accomplished' for obvious reasons," Ozzie said in an interview after his. "But I'm really pleased with where things are."
It's, to be sure. But Ozzie says Microsoft has changed in ways he could not have imagined. In particular, Ozzie points to Windows Azure--Microsoft's operating system in the clouds. Rather than just offer a set of services to move today's computing programs to remote servers, Ozzie says Azure is designed to handle the applications of tomorrow.
"When we began developing Azure, we developed it more or less with a clean sheet of paper saying, 'What will the operating environment look like for the next 30 years?' Ozzie said. "If you look at VMware or (Amazon's) EC2, what it really is--and I mean to be saying this respectfully--but it's more or less a (virtual machine) hosting environment. It's not a transformational computing environment."
In a lengthy interview with CNET, Ozzie also talked about lessons Microsoft learned from the recent Sidekick outage as well as why people are wrong to count Microsoft out of the smartphone race.
Here is an edited transcript:
Question: From your perspective, where would you say Microsoft is in terms of making the kinds of shifts you talked about in 2005? What is different than you thought it might be?
Ozzie: You know, when I wrote the memo, I really didn't have a crisp plan in terms of how we're going to accomplish it. And I will never, ever, utter the words "mission accomplished" for obvious reasons. But I'm really pleased with where things are. I mean, I think we have a lot of software yet to deliver, but out at the end user perspective, the notion of Office being across phone, Web, and PC, kind of re-pivoting the experience around productivity as opposed to the device, I'm really happy about [that].
I thought users would be more ready for it by this point in time than I think people really are. I don't think in our minds yet we've yet found, quote unquote, the desktop for the Web in terms of our own personal stuff. It's kind of still scattered out there on the Web.
I didn't think that the cloud computing thing--the back-end side--would take off as much as it has. There wasn't as much about that in the memo, but at that same time, you'd probably be amused to see some of the PowerPoint decks that I was shopping around internally at the time with these big pictures of hydroelectric dams and all these things saying there's going to be this recentralization that happens at the back end of computing, but I didn't know how it was going to pan out.
You announced that Azure is going into production January 1. Is the code changing significantly between now and then, or is that just when the billing mechanisms kick in?
Ozzie: What happens is--and this is all just really difficult to explain to people--but we've rolled out big, new data centers. The community technology preview is on a certain sets of servers. Some of those people may or may not opt to become production customers. Getting their things migrated from one set of systems to the other, it's just internal logistics. So, no, the code doesn't change a whole lot, it's more operational processes. And we really don't want to start charging people until we at least have one billing cycle of knowing that everything is right.
You mentioned moving people from one set of servers to another and immediately I hear in the back of my head "Sidekick." Obviously, the architecture is totally different. But can you talk about what you took away from that [
Ozzie: There are a lot of lessons to be learned. Let me just preface this by saying it's inappropriate for me to go deeply into it not just for legal aspects and things like that, but because they're T-Mobile's customers, not ours. T-Mobile is our customer. But let me just speak at the abstract level.
There are lessons to be learned in terms of how acquisitions are dealt with. I know that's a non-obvious conclusion, but basically when you're building your own services and when you're building services from scratch, you have a certain understanding because of the people who were involved in that or whatever--of how this thing relates to that thing. When you bring in a company, you tend to think of things differently. And so there were some lessons to be learned there. There were lessons that we didn't learn, (areas where) we know better and I'll just say we weren't using best practices in certain areas.
The biggest lesson is something that I shouldn't have had to learn, and I'll tell you why. In Groove, I took, for the time, a very contrarian view of, no, it's got to be all at the edge. Nothing at the center, it's all peer-to-peer distributed. Then we--and I mean including me--have kind of swung the pendulum to appliance-based computing that's Web-centric, where the truth is in the cloud, so to speak.
One of the fascinating things about the Sidekick recovery process was how wonderful it was that data is also on the devices, because when your confidence level drops in one copy of the data and you have another one, it's really handy. So knowing to treat peer computing and centralized computing are both good, they're both very, very good.
You talked about the cloud as being early days. And I'm curious, there are some folks that have been playing in the space for a while, you know, SalesForce and Amazon and even Google to an extent. What do you feel Microsoft is offering in the cloud that competitors aren't?
Ozzie: When we began developing Azure, we developed it more or less with a clean sheet of paper saying, "What will the operating environment look like for the next 30 years?" If the servers like Linux and Windows NT-based systems and Mac OS, if these are all based on things that were built when I was in school, what's the next one going to look like? That's the most significant advantage.
If you look at VMware or [Amazon's] EC2, what it really is--and I mean to be saying this respectfully--but it's more or less a [virtual machine] hosting environment. It's not a transformational computing environment. All programs in the future will be written in a way that there is no single point of failure. There's no one server that can die and take down the service. And unless you write your applications for a programming model that's inherently parallel, you don't get to that point. And so, yes, we support the same kind of mode that the EC2 or VMware will do where you can take a VM and put it up there, but the reality is you don't get the benefit of cloud unless you use this other thing.
You actually had to go back and add that in. One of the things you talked about today was to take a virtual machine and put it up on Azure.
Ozzie: That's a very good observation. Last year, we introduced, I guess I'll say [something that was] a little too far ahead and we had to back into the present. But I'm extremely pleased about [adding the virtual machine ability] because anytime someone starts playing with [Azure] and they start to get a taste for what it's really like, then you really say, oh, I get it. Now I know how to design the software for that next generation.
You talked about "three screens and a cloud" as a pretty consistent refrain for Microsoft. But we're still not hearing as much about some of those screens, particularly on the mobile side. You mentioned in the spring we're going to hear sort of about the next-generation platform?
A lot of people are saying, you know, Microsoft and the phone--it's been way too long, game over. Why is that not the case?
Ozzie: I think it makes for good copy to take an extreme position that someone is dead or alive or this or that. Yes, iPhone has a lot of momentum, unquestionably. But I think the phenomenon we're in right now is the app phone. And if you look at the depth of apps that are on these phones, they're not very deep. It's not like Office or AutoCAD, where there are just thousands of man years that have gone into developing these apps. They're relatively thin apps that are companions to some service.
And I think if you look at anyone who's building an app phone--whether it's Palm, Google with Android, RIM--ultimately, all the apps that people want will be on all the phones. They're relatively straight porting efforts. I think people are imagining some kind of a barrier to entry, at least from an app perspective that I don't believe is there.
The biggest barrier to entry is: is it a phone that people want to use? And is it a phone that carriers want to sell and people have to measure us based on what we produce. But I don't believe that there's an app barrier.
This year, it seems like you guys have made a conscious choice to focus on Azure and not on some of the more finished services that live one or two layers up. Are you still pursuing the sort of Live Mesh and the Live Platform layers?
Live Mesh, as a specific case in point, after we got to a certain point in the beta, we said, okay, how are we going to get this to scale from instead of a million or two million people to hundreds of millions of people? So the team and the technology was put into Windows Live and so even though I'm not making a product announcement, when you look at the next version of the Live services that are downloaded to your desktop, I think you'll see the contribution that the Mesh technologies and the Live platform had to that.
In terms of high-level services, no, we're still concentrating [on them]. You know, we still have a very big focus on the Web apps. I think you probably won't hear a lot about that at PDC, but you'll hear some more about that as Office comes more into a broader beta.
Between Pinpoint and Windows Marketplace, Windows Mobile Marketplace, Zune Marketplace--you guys have a lot of marketplaces.
Ozzie: It'll be converging down to two, one for consumers and one for IT and developers. Yes, it's a big company, yes, we have many ways to sell, but ultimately, there should be one place for consumers to buy things online, you should have one shopping cart across this and that. That doesn't necessarily mean one [user interface] to the marketplace because when you're in Xbox, you want to see it through Xbox. When you're on a phone, you want to see it through the phone.
On the PC, I'm still not actually convinced what the right thing is. When you're on a PC, do you want to see the marketplace through the Web or through a client? You know, I can kind of see both. I mean, look at the Zune marketplace, people like being able to buy it through a media-oriented marketplace, but if you were buying apps, it's not really clear. But in any case, there's one marketplace back end that is syndicatable into multiple front ends for the consumer and for the enterprise/IT, and what we were talking about today was really the enterprise/IT one.
It struck me that today, a lot of the story about the cloud has been that it's great for load balancing, it's great for sort of having predictable investment in IT, but there hasn't been as much about what are the benefits when your app is running in the cloud. It sounds like the new project code-named Dallas could be an example of one of those things where you can build a type of application that you couldn't build on premise because you're using someone else's data.
Ozzie: It is the right way of thinking about it. What we're basically trying to say is by agreeing to get together in a certain way, by agreeing on certain guardrails on the road that we'll all drive on, there can be benefits. Right now, there are many pieces of public data, there are lots of commercial data providers and each one has a different kind of a licensing mechanism. Some license by developer, some license by customer, some license by individual user. There are just lots of different terms. And a lot of the big benefits in the data that's out there are what happens when you join them, when you bring them together. And I believe that there's going to be a lot of potential in this.
Will we see Microsoft be kind of one of those first and best customers, bringing a lot of its data and making it available ?
Ozzie: I think the biggest set of data that you'll see us take in many directions is maps. It's the most obvious from a consumer's perspective. You can layer upon it quite nicely. You can layer both apps and other forms of data on it quite nicely.
What are some of the things that people have developed on Azure? Are there any areas of types of applications that have particularly surprised you?
Ozzie: I'm not sure if you noticed some weeks ago, Qi Lu was at Web 2.0 and he announced this Twitter on Bing feature? That is on Azure. And it's one of the most fascinating stories in terms of agility.
A number of people from across the company looked at this thing and said, "Wow, if we had the Twitter fire hose, what could we do with it? Let's start experimenting." And this other lab said, "Oh, well I already know what to do, you actually have access to the fire hose? How could we ever get enough machines put together in time?"
And just in a matter of weeks, you know, this app just came together, people came together, and we had this thing live. And the number from the virtual machines that are processing the incoming feeds, it's fairly astounding. Since that time, other experiments involving 2,000 machines here, 3,000 machines there, are just popping up because people haven't conceptualized what would it be like to have that kind of resources at your disposal.
Are these the kinds of data feeds we're going to have in the future? I mean, Twitter, you have this tremendous data feed, but you can't take in everything, at least not over an extended period of time right now.
Ozzie: In late '05, I guess it was, when I wrote that last memo, I had a theme that I was kind of talking about internally about moving to the cloud experiences and the back end. These days, I'm basically asking people the question: What if everything was recorded, everything? You are recording in your pen there. Some phones have the capability now--or maybe they're just prototypes that we've got--but measure barometric pressure, measure temperature.
Obviously, there are accelerometers. If you can measure everything and you have this aggregated data, what can you then do with it? And I think just getting people to experiment with it will bring us to places that we haven't known before. People concentrate so much on the scary aspect of privacy related to advertising base uses of it, but there are other uses.
From a health perspective, there are many things that I could measure about myself that would be of value to me and no one else, but we still aren't building those apps. It's just too hard to gather all these things.
When you kind of look at where you are, what are the gates to getting where you want as fast as possible? Is it still a matter of evangelizing inside the company? How much is it still a challenge that Microsoft is such a big company that is divided into product teams responsible for the here and now? What are the things that are sort of the biggest gates?
Ozzie: I would say the biggest gate is the same gate it's been for several years, but it's trending in distinctly the right direction, which is prioritization. It's just simply there are a lot of opportunities, there are a lot of different directions that we could go. And left unchecked, every time you do something new, it causes more complexity.
One of the positive side effects, if you will, of the economic downturn is the fact that we've all been forced to make the hard choices.