Showing posts with label longhorn. Show all posts
Showing posts with label longhorn. Show all posts

Monday, November 14, 2005

Hypervisor: Status update

Back in a post from earlier this year, I talked about some of my thoughts on Microsoft’s hypervisor layer of technology being built into Longhorn for virtualization. At the time there was really very little to go on publicly, beyond a commitment to improve virtualization technology in Windows products, and some older research material on virtualization technology in general.

Since then, it looks like quite a bit has happened. First of all, this webcast is really worth checking out. If you’re specifically interested in the hypervisor layer, jump to minute 43.15 and start listening. But to hit the basics, it looks like I was at least in the right ball-park with my with my initial take, or if not my initial take, then at least with some of my wish-list items.

My understanding is that the microkernelized hypervisor is a thin layer of software that sits just below the host operating system to interact with the hardware. This is notably different then the way Virtual Server 2005 exists today, where the vmm.sys sits architecturally at the same layer as the kernel of the host OS. From a functional standpoint, this should improve reliability and reduce or eliminate the potential for something happening on the “host” to bring down the hypervisor layer, and thus crash all guest operating systems

So what’s happening at the hypervisor layer? Well, some very basic core functionality, with an emphasis on dependability; which makes me wonder if we’ll see something like Microsoft’s Singularity doing work here. If you haven’t heard of Singularity, check out this link for a PDF of material related to the effort. The short of it, is that it’s a research project which has a goal of demonstrating practical new technologies which improve dependability. Sounds just like what we’d want in the hypervisor layer.

Step back for a minute and think about the implications of this. We’ve already heard that Longhorn is going to include something called the “Windows Server Core”. Which is really taking the concept of “roles” that were introduced in Windows Server 2003, and extending that concept out. In other words, we’ll be able to deploy a “Windows Server Core” with say, IIS. What that means is that the server core will be deployed to the bare-metal, as well as TCP/IP, and IIS. We won’t get a GUI, we won’t get unnecessary services, we won’t get Internet Explorer, accessories, RRAS, etc. We’ll get exactly what we need.

So in the datacenter Virtual Server 2005 already matters. We’re continuing to see server consolidation projects where low-utilization servers are being virtualized, saving all kinds of resources (hardware, power, physical space, etc.). Longhorn and hypervisor hold the promise of taking the advancements achieved in Virtual PC and Virtual Server, and really improving them in terms of dependability, and functionality.

Thursday, June 09, 2005

Longhorn CLI: Monad shell (MSH)

More news on Longhorn! It sounds like the Monad Shell (MSH) is going to be replacing, or at least supplanting WMI for scripting administrative functions. It only makes sense to move to something based on the .NET framework, and from the looks of things MSH is going to be more on par with something akin to bash. While I recall hearing about Monad a while back, it’s good to hear that things are progressing on the CLI front.