Microsoft"s Jim Allchin talks up Longhorn

Von Carol Sliwa

Microsoft Corp. is making preparations to hand out a preview copy of its next major Windows release, code-named Longhorn, at the Windows Hardware Engineering Conference later this month in Seattle. The first Longhorn beta release will follow this summer, and while no date has been set for a second beta release, Microsoft is pledging to ship the desktop version of the operating system by late 2006. The server operating system will follow in 2007.

Jim Allchin, group vice president of platforms at Microsoft, said the company will put "massive marketing" behind the new operating system, which will be its main platform for the next 10 years. Some key areas of focus for Microsoft in developing Longhorn were improving security and safety, adding new capabilities to "visualize and organize" information in ways that move beyond traditional searches, and reducing operational costs.

Excerpts from Part 1 of the interview with Allchin follow. Part 2 will be posted online on Monday.

What new features in Longhorn are tailored for IT shops? In business, we want to be able to manage the images that people are creating for deployment. Today, it"s very complicated for them. They have to build images for different locales around the world because of different languages. They have to build images that are different depending on the type of hardware that they"re deploying it to. All those add cost. We"re trying to do a re-engineer of that to make that much simpler.... That"s something we"ve been asked for from the OEMs, along with the corporations.

Another example: We"re going to drop the number of reboots. We"ll do ad hoc patching. There"s a whole set of things we"re doing to try to keep the system to where availability is higher.

Do you have a goal for continuous uptime? We do, but I"m not going to quote it.

How will role-based computing work in the Longhorn server operating system? Our focus is to take the experience thinking and tie it to roles that the server is in. This is a Web server. This is a messaging system. This is an AD certificate system. Once we know what role it is, we want to thoughtfully understand the task that a particular administrator might do on that particular type of machine. You don"t want them jumping all over different tools to do their job function.

For example, if you"re a Web server administrator, what do you typically do? Are there cycles you go through? How do you do maintenance? We"re trying to think through what happens there.... We"ve created roles. You check that role, and everything you need for that role is there. You don"t have to think.... You say this is a SharePoint server. That"s all that"s going to be there. You"re not going to have the rest of the Active Directory bits loaded there.

It"s like a Swiss Army knife, only instead of having the rest of the blades there, which might get in your way, you basically say I want this blade and the rest of the blades fall away.

Does it also mean stripping away everything that"s superfluous and getting back to some sort of core plus a set of dedicated features? Once you decide that that"s what this server is, then that"s what runs in that box. We provide you the ability to have one role or n roles all in a single box. We did it a little bit in (Windows) 2000 (Server), more in (Windows Server) 2003, and we"re just taking it to the next step here. On the box, you"re only going to run the stuff you want to run.

But there won"t be separate server SKUs based on roles? Not on a separate SKU plan.... Our idea is not to splinter off and have separate SKUs, because we think that would be more complexity.

Will it be similar to a Linux environment, where you say I want these parts and I don"t want the rest of it and, lo and behold, you create an instance of the operating system that is exactly what you want? Yes, with a caveat. With Linux, they take it all away. Here"s the code. Just compile out the stuff you don"t want. We"re trying to make it so you just do a check box, and the stuff just isn"t there. It"s still available on the disk, if you want it, but it"s not in the image that you"re using on your machine.

Which Longhorn features will help most in your competition against Linux? We are working on partitioning. And that"s the ability to add processors and add memory while the system"s running. There"s a whole set of availability. The ability for fewer reboots. Componentization, I think, will be appreciated as well -- and the role-based approach.

To what extent will virtualization capabilities be built into the operating system? I hope everyone understands that virtualization is sort of a native part of an operating system from ground zero. We"ve virtualized the CPU to give processing. That"s what we do. And we"ve virtualized memory. That"s what virtual memory comes from. So all that"s happening now is, as the hardware progresses with more capability for virtualization, the OS is going to take advantage of it. Today we have stuff for products that are there because the hardware really doesn"t do everything we need it to do. But as the hardware does do it, it"s just a natural for the OS to support it.

On our current path, (there"s) some isolation that we do in Longhorn. Virtualization is not planned for Longhorn; well that"s not true -- some parts of it we are considering. But we won"t make it because the hardware won"t be ready. For example, virtualizing the I/O, it"s not there from the hardware, and that"s something that we would really like. So we will progressively extend the virtualization in the OS to take advantage of the hardware virtualization that"s there.

Will you take features from your Virtual Server product and fold them into the operating system? What we want to do is take more advantage of the hardware as we move ahead.

Will the Virtual Server product eventually go away? I can envision the path that there would be no Virtual Server product at some point. However, I could also envision the path that says there"s a thin hypervisor-level system and that there"s a separate virtualization stack that is sold separately. That"s also a possibility. So I don"t know. I think the world will evolve here, and I don"t think that necessarily anybody"s products today will necessarily stay the same in the future, because the hardware"s going to change this.

What"s up with WinFS, the storage subsystem that was dropped from Longhorn? The way we"re envisioning it right now, and it"s very early, subject to change, is that it would be doing beta about the time that the client release is. And when we would ship, we would make it available for our (independent software vendors) to start using, and then in a future release, the OS would include it.

Would an IT shop likely use it? It"s too early. And I think the proof will be in the pudding -- when we start showing it, and people say, "OK, do I think this adds value or not?" We obviously think it does, but the proof will be in the pudding.