run over Azure as well. They are very complementary. We are in the same data centers, we have a shared infrastructure. Going forward, you have the option to use one or the other.
X: So you’re saying that Microsoft will be eating its own dog food—it will use Azure in the data centers where the cloud versions of the Office application suites are running?
YK: At the moment they are not, but they will be run over Azure. We haven’t said when.
X: Beyond the fact that Azure is customized for the Windows environment, how would you describe the main differences between Windows Azure and competing cloud platforms such as Amazon’s EC2 compute infrastructure and S3 storage services? It sounds like the model-based approach that you’re talking about is one of them.
YK: Let’s start with the basic value proposition of the cloud. It’s two-fold. One is cost—reduced cost, both operational costs and capital expenses. The other has to do with agility, the ability to respond to business needs very quickly. We believe strongly that to get the value proposition out of the cloud, your application set has to evolve to meet the cloud. You have to meet it halfway, in a sense. This model-driven cloud application architecture is not new; people have been doing service-based architecture for a while now. What we are doing is making it easier to write such applications, and we are pushing it as a way to write applications in general, regardless of whether you run them in our public cloud or somewhere else.
So we have invested heavily in mode-based programming tools to make it easy to make applications cloud-ready. Once they are ready, that is our big differentiation point. If you look at what others are doing, they’re focusing just on the speed of a given virtual machine or a given service. We want to have a platform that is open to everybody, where it is easy to write applications and where applications benefit from being written to that model. You mentioned Amazon by name. We have a high level of automation and software patching of the operating system built in from day one, for the same price, unlike others.
X: Given that Microsoft is a software development company first and foremost, employing thousands of software engineers, it wouldn’t be surprising if the cloud platform you have built, namely Windows Azure, was basically created by and for developers. Is that a fair characterization?
YK: You are almost correct, but let me say that in my own words. We have two audiences in mind: the developers and the operators of the applications. So basically, yes, we wanted to make developers’ lives very easy, and we wanted to enable them to do application development really easily. But it’s also for the operators of applications. How did we arrive at this model? To be honest, we have been in the online business for more than 10 years now. We have a long list, a few hundred properties and services online. We have studied internally how we built those services, and we are leveraging the models and the work that people have done and have upgraded them to the cloud.
X: Come February, when Windows Azure goes fully commercial, what types of applications do you think customers will actually run on it?
YK: There are some things that are easier to move to the cloud than others. There are extranet-like applications, where you have to cooperate with multiple entities; collaboration; federation; Web-based applications. I wouldn’t say these are “no-brainers,” but they are the first ones you move to the cloud. There is also a class of applications that require heavy-duty computation, number crunching, and high-performance computing; those also run well in the cloud. Then there are applications that may have a Web-facing piece that lives nicely in the cloud, but still have details or components back in the enterprise. For example, your heavy-duty database containing your employee data may be on-premises, but the rest of your ERP or internal accounting software may be in the cloud.
Some things may not be movable to the cloud, because of security or compliance, or because other apps need to use it. Here at Microsoft, we run a big auction once a year where employees can go bid on items. For one month a year, that requires huge capacity, because everyone is using it, but for the rest of the year it is not running at all. That’s an app we ran this year on Windows Azure. At PDC, we demonstrated how such an application can be securely connected back to the enterprise.
In a nutshell, not all workers will move to the cloud tomorrow, and those who move, initially, will be working on websites, extranets, collaboration, and other areas.
X: When we started planning our Cloud3 event, we initially wanted to take some of the more longstanding questions about the security and reliability of clouds off the table and treat them as largely solved, but we got feedback from many people that those are still important, open questions. So how do you answer questions about the security and reliability of Windows Azure?
YK: We believe there are workloads that are very appropriate to the cloud today, and I mentioned examples. There are others that may not move into the public cloud anytime soon, and for those, we would say we have a place for you to run those, within hybrid models like a dedicated cloud hosted by somebody else. We are saying we don’t believe everything will move to the cloud tomorrow, nor do we want it to. Rather, we think you should have options and flexibility of choice.
Others may argue that everything will eventually move to the cloud, and it’s only a matter of time. But frankly we don’t see it this way. There are compliance and regulatory issues and national boundaries that may require you to run things in their current places. But we will beef up the security of public clouds under the same programming model. That is why I keep harping on the point about models—if you write in a certain way, it becomes apparent which pieces should run in the cloud and which pieces should be private. My PDC presentation was on exactly that subject.
