Q&A: Valentine explains Microsoft’s management push

We discuss Microsoft’s new multi-year management initiative with Brian Valentine, senior vice president of the Windows division.

Microsoft this week took the wraps off an ambitious multi-year plan for developing a self-managing computing environment. The plan begins with Windows Server 2003 and updates to System Management Server (SMS) and Microsoft Operations Manager (MOM). It continues with new technologies such as System Definition Model (SDM), an XML technology used by servers and applications to exchange management data, and Automated Deployment Services (ADS), technology for deploying software. Network World Senior Editor John Fontana sat down with Microsoft’s Brian Valentine, senior vice president of the Windows division, to talk about the company’s management plans (discuss the plans in our Microsoft management forum).

Describe Microsoft’s shift in management strategy.

There is a lot of momentum in the industry right now around general-purpose compute environments that then become workload-specific dynamically. Today everything is silo-based. You have these sets of applications that run on these sets of machines and you treat it as a silo application, and it is not really an integrated environment. Well, our strategy really turns that on its side and creates a computing environment that is an enterprise service that can be dynamically reconfigured to solve whatever business problem you want to solve at the moment. But it also says that to do it right, you have to be able to capture the entire lifecycle of the application, from architecture design all the way to end of lifecycle. You have to really take that and build it into the platform.

So, break down the architecture.

If you look at the management architecture in Windows today, it is not the world’s best, right? The event log – we know about that. The registry – we know about that. MSI [Microsoft Installer] – we know about that. WMI [Windows Management Instrumentation] is there and is probably the best part of the management architecture, but it is only a small piece. And then you take some of the services on top of that. I think Active Directory is pretty good. But some of the security infrastructure stuff is still clunky and hard to use. So there are a lot of things that are there that can be used as building blocks, but are they really truly fundamental architectural building blocks that you want to go forward with over the next 20 years? No. But the needed building blocks cannot be built by a third-party ISV to truly deliver on the vision of a managed operating system platform. So, taking a lot of what has been provided in the management industry so far, and integrating in the platform elements in the operating system is a big bold statement.

Does this initiative compare in scope to Trustworthy Computing in that it touches nearly everything?

Yes, everything in the enterprise. It is a bigger effort from that perspective. When you look at the benefit, I think it actually has a bigger benefit than Trustworthy Computing. Security is a huge issue. You’ve got to provide a secure solution, but that is pretty vertical and one-dimensional. This is a platform play across a lot of things so the impact of it can be a lot bigger. With this we can advance computing infrastructures dramatically, where security doesn’t do that. So it is different from that perspective, but this is a pretty big undertaking.

You haven’t talked about integration with other platforms. When you talk about the data center you need integration.

Our view is that there will be heterogeneous environments for quite some time – forever. So we want to have open interfaces, we want to have open protocols. People can go develop whatever infrastructure they want to develop on other platforms, and they can decide at that point if they want it to play within our environment or not.

How are they going to make it play in your environment?

SDM is an open thing. You can have multiple producers of it and multiple consumers of it. So it will all just play within that model. Is that infrastructure exclusive to Windows? No. There will connection points to that infrastructure. There has to be. We can’t go into a customer and say it is a Microsoft-only solution.

You have not mentioned Application Center or the acquisition of Connectix virtual server software. Where does all that fit?

The scenario that we wanted to resolve with Connectix is, when we first delivered Windows 2000 it meant that customers had to port apps, and many of those apps are legacy apps and it made no sense for them to crack open the code to port it. So they were stuck on NT platforms. So the virtual machine solution is a great solution for that. But it doesn’t solve dynamic partitioning across a single operating system version. We are continuing to invest in those spaces.

So does Connectix fit under the DSI model?

Yes, it could. But it is a very small piece of it.

And Application Center?

We had a project that came out of our research group a few years ago internally called BIG – ‘billions of interconnected gizmos.’ It was all about scaled-out computing and dynamic application development. It became Dynamic Systems Initiative (DSI), basically. We have been working on that for three years now. When you think about developing an application, that is a tools issue. When you think about deploying the application and managing the application, most of that is either operating system or platform elements. Application Center’s vision when it was started was to be a distribution and management environment for distributed applications. Now, that gets consumed by the operating system in the plumbing aspects – how you distribute images and how you manage those images. The new Automated Deployment Services is 20% to 25% of what App Center does anyway. Then the management part of App Center is part of the management vision and gets integrated some into the future System Center product. So it did not make sense to keep App Center as a single product. We will support the current customers and move them forward. But the way they move forward is not necessarily to buy App Center version 4 in the future. You buy System Center and the operating system takes over the other duties.

What would you say to people who say Microsoft’s three- to five-year timetable for delivery of a self-managing infrastructure is pretty ambitious?

The entire vision of absolute simplicity, flexibility, and automation in the data center could be a 10-year initiative; it could be a 20-year initiative. The thing about Microsoft is that when we decide to go down a path like this we are pretty persistent. We’ll put in the appropriate resources and we will learn as we go along. When you look at the architecture there are a lot of fundamental building blocks, and we are starting down that path with Windows Server 2003 and Automated Deployment Services and some of the additional things that you will see this year. The next version of Visual Studio will produce SDMs, and the new application products we build will consume those SDMs. The operating system management infrastructure will get dramatically improved over the next couple of years from being event log, registry and MSI, into something that is much more oriented toward application isolation and installation, and configuration management and software distribution. The operating system will know how to update itself and also provide a platform for software update and installation. Those are all the things that we can overlay on top of Windows 2003 and deliver with the Longhorn wave that allow the ecosystem to start going hard.

If I’m IT, where do I begin to understand this?

I would understand it from three aspects. One is what the Visual Studio guys are saying. That is the design-time element. Second is what kind of platform elements are being delivered over the next year with Windows 2003 and some of the stuff that we overlay on top of that, such as Automated Deployment Services. Also, if you are an SMS or MOM customer today, what does your roadmap look like for the next two, three, four years? And then as we go more public with Longhorn and that wave of stuff, stay in tune with that. But the place that it starts is with Windows 2003 Server, because what we have done in that release is to make sure that the architecture supports the overlays as they come.