Ray Ozzie Explains Cloud Computing For The Enterprise

How should IT departments prepare for Microsoft's forthcoming Azure cloud computing services? That was what we wanted to know when InformationWeek sat down with Microsoft chief software architect Ray Ozzie at the company's Professional Developers Conference this week in Los Angeles.

On Oct. 27, Ozzie unveiled the Windows Azure "service foundation" and Azure Services Platform, which together are Microsoft's answer to Amazon Web Services, Google App Engine, and other on-demand computing services. Azure is still an early stage services platform; it's available only in a limited technology preview for developers, and Microsoft didn't disclose when Azure services would be generally available. Our best guess is the second half of 2009 or later.

Azure services will be hosted in Microsoft's data centers, giving businesses an alternative to managing their own servers, databases, and storage systems for some applications. Ozzie suggests that Azure could result in lower IT costs for businesses, but Microsoft hasn't disclosed pricing. We interviewed Ozzie at the Los Angeles Convention Center.

INFORMATIONWEEK: We're glad to be able to finally start talking about Microsoft's cloud strategy.

OZZIE: I've been talking about it in the abstract for so long, even though they are technology previews, having code and demonstrations to back up the vision is really great. It's exciting.

INFORMATIONWEEK: It's not as though all the work is behind you.

OZZIE: No, it's just beginning, the grand opening.

INFORMATIONWEEK: We want to delve into this from the point of view of the IT department. What do enterprises need to do to get ready for Azure and cloud computing?

OZZIE: The fact that we're beginning to talk to developers about this means that we can begin the conversation with IT about what all this means over time. That conversation is nuanced, and it relates to virtualization and "private clouds," using this kind of technology within their own data centers incrementally to save money and operate more efficiently. It should begin to help them analyze the kinds of things they have on their machines to separate infrastructure that doesn't necessarily add unique business value from the systems that they really believe are unique to their business and that they need control over.

If there's an opportunity to run the infrastructure pieces in the cloud, they should know which ones they're willing to try first. Microsoft has online offerings [CRM, Exchange, Office Communication Server, SharePoint] that are actually going to be the first touch point of most of our enterprise customers with the cloud. Azure is great for leading edge developers, people who really want to get their skills up to date in terms of where things are going, but for enterprise IT, Exchange Online and SharePoint Online are probably right up front.

My recommendation to every enterprise worldwide, to any enterprise that has Active Directory, would be to put a small number of seats of Exchange or SharePoint online, because in doing so you immediately learn how to connect your directory up into the cloud. You start to deal with a management console that lets you provision services to users. You'll start to confront your own issues of up time. You'll start to frame what kind of a pipe you'll need into the Internet. It will give you the ability to start knowing what you don't know.

There's a tremendous amount of promise once they get to that point, because once they get comfortable with a small number of seats, then it's a very quick decision, it's a group policy change, and then suddenly they've got lots more users on it, and their users will never really be aware of that switchover.

For the enterprise and for the enterprise developer, the classic VAR who builds apps that serve the enterprise, the biggest news was that once an enterprise brings online or online-branded things into their company, third parties will be able to be provisioned exactly the same way. Today it's very difficult for a niche solution provider to provide a service to an enterprise. It's not like software; a VAR will bring in a disk, fire up a server, connect it to Active Directory, and use standard group policy to provision it to users.

For services, it's not that easy. Even with all the promise of easy on and all this, in order to provision it to users internally, what are you going to do, give each person a new ID for that service and a new ID for that other service? We aren't in a world right now where federation is broadly used. And we've created in our online properties one-click federation between the enterprise and our Live ID ecosystem so that now business developers can hook into that and serve a world of enterprises that they might not have been able to do otherwise.

INFORMATIONWEEK: There's a list of concerns that CIOs have -- security, governance, data integration between one service and another service. How do you respond?

OZZIE: There are even more. In composite services, there are issues of latency when something is in one data center and another piece is in another. This is why it's a good thing to begin dipping your toe in the water. It's not a panacea, and it's not something that I think people should fear. It's a huge opportunity for cost savings and better scalability, a variety of things, but you have to grow to become comfortable with it.

There are certain things that I'm extremely confident of that most enterprises will get over, things that I used to be worried about [such as] information confidentiality issues. At least forMicrosoft (NSDQ: MSFT), we've been under extreme scrutiny for a number of years in terms of our data handling practices, and we're very transparent. We can make it very clear and make most enterprises comfortable with the fact that their data will be handled well. We are another third party, though, so there are law enforcement issues that enterprise should be aware of.

But the flip side of the coin is the promise that, in our cloud, we can bring services to many enterprises that they never would have deployed themselves. We can do auditing functions and information rights management functions in the cloud that might take some configuration and management that many smaller or medium-sized companies wouldn't have bothered to deploy. We can bridge their communication systems to mobile devices and do a lot of things that might have required more investment of IT on premises.

INFORMATIONWEEK: You mentioned cost-savings. Is cloud computing really cheaper?

OZZIE: That's not the single reason that somebody should do it. If somebody wants to expense something rather than capitalize it, they can do that through leasing and things like that. There are some financial reasons, but that's not the primary reason. The primary reason is really if you're going to have people, headcount, employees, do you want to be training those employees in how to manage F5 load balancers, how to doCisco (NSDQ: CSCO) IOS, and debug high-scale network problems? Do you want to have the coordination between the operations personnel and the development personnel in order to ensure the kind of availability that you need? We can bring a lot of that experience to bear.

Ten years from now we're going to look back at this era and wonder how we did without this other kind of computer in the cloud, because for so many things that will be the default place that we'll just run it. We won't run everything there, and if you didn't get the message yesterday, we keep saying the power of choice, the power of choice. One of the things that we are trying to communicate is that enterprises have a broad variety of different needs, and we're going to offer them on-premises with all the controls that they need. If you're a government that has a network that's disconnected from the Internet, you're going to want those server-based products. If you want to control the versioning and exactly when things are changed for your employee base, we'll give you all the ability to do that. But if you want to take advantage of the service-deployed way of doing those same things, we'll also deliver that.

INFORMATIONWEEK: When will enterprises be able to take advantage of Windows Azure?

OZZIE: I can't tell you exactly when. I can tell you it's not going to be in '08, but over the course of '09 we will pay attention to what these developers [at PDC] tell us. There are a lot of capabilities that are in Azure that are not unlocked, and there are capabilities that we have not yet implemented. Based on what these folks tell us, we will prioritize things differently. So, I'm not trying to be evasive; I'm simply saying that we just don't know.

That said, it will probably be first utilized as the Web tier of things they do, the Web front-ends. The simple, stateless kinds of machines that they run will be the first things that go up into the cloud in terms of them coding against Azure directly. They'll be using it indirectly through online, because the online properties will use it, the Mesh stuff will use it, a lot of things will use it.

There are other ways, too, such as Exchange hosted services. Those things are all being ported over to Azure, so that when somebody subscribes to message archiving, the message archives are going into SQL in the cloud, and so on. But directly it's probably going to be the simpler front-ends.

Over time they will then begin to experiment with Web enterprise applications and build from there. It will be generally be new applications, not existing ones.

INFORMATIONWEEK: Can you give us an idea of the amount of internal resources that are going to this development project and a little bit about how much more work is to be done?

OZZIE: That's an interesting question, because underlying that question is a value judgment of what that number really means. To me, less is better, so I don't know how you would judge it. All I can say is this: From a services transformation perspective basically everyone in the company, all of our products are in some way, shape, or form being transformed by the fact that there is a services component to it. The SQL group is fairly impacted. Obviously the Live group, they've always been impacted. They are a service at birth, but that group has grown because of the appearance of things like Mesh, which are platform technologies.

I don't know if I should give you an order of magnitude. The Windows Azure project has grown like most new projects should grow, and I'll give it to you in an abstract way. Any new effort, if you don't start it in the range of, I'll say, 10 to 50 people, you're throwing too many people on the project too soon. You're not going to create something that's very coherent. So you get just A players, the smallest number of A players you can, and then you start broadening out. We're in the broadening out phase right now.

But it's a serious project. It's got significant capital expenditures for the data center infrastructure necessary to host these things. It's an operating system.

INFORMATIONWEEK: The Office Web application demonstrated today looked rudimentary. Is that an indication that it will be relatively rudimentary in what they can do?

OZZIE: Tell me in what ways you say rudimentary?

INFORMATIONWEEK: I don't get a sense of the scope of collaboration that would be capable, real time collaboration across the Web and the client app, if it's going to be different in the different applications, if there are deeper features than just being able to see what somebody else is typing.

OZZIE: Well, I'll tell you how the project was approached, without saying what state they're in. They are being developed by the core group that's developing Office. It's not an adjunct, not a little thing off to the side. It's treated as a part of the core, and, I can say this, it's the same source code. A lot of it is the same source code.

The issue is we put a primary pivot on the scenario that you and I need to work together on something. I can be using the Web; you don't need to know. That design decision very early on dictated what pieces of code really had to be the same code that's operating in the service and on the client and so on.

And, just like there are PC-native features that only exist on the PC, there are Web-native features that only exist on the Web. I would say we're probably less mature on that. There are some, but we're less mature on that because our primary pivot was round-tripping.