[Beowulf] small-footprint MS WIn "MinWin"
James.P.Lux at jpl.nasa.gov
Mon Oct 22 13:15:50 PDT 2007
At 10:14 AM 10/22/2007, Robert G. Brown wrote:
>On Mon, 22 Oct 2007, Peter St. John wrote:
>>This from Slashdot http://slashdot.org/article.pl?sid=07/10/18/236233 :
>>"Some small but significant details of the next major release of Windows
>>have emerged via a presentation at the University of
>>Microsoft engineer Eric Traut. His presentation focuses on an internal
>>project called "MinWin," designed to optimize the Windows kernel to a
>>minimum footprint, and for [sic] which will be the basis for the Windows 7
>>I thought this was interesting on account of the brief discussion a few
>>months ago about the footprint of MSWin vis-a-vis clustering.
>It's also interesting in other ways. It sounds like Windows is evolving
>towards an ever more Unix-like design.
I don't know that Windows (at least since NT) isn't actually already
a small kernel surrounded by (dare I say "embraced and extended by")
a lot of utilities. Sure, they're not command line utilities with
cryptic 2 letter names and a man page full of switches. However, an
awful lot of what people talk about as "Windows" isn't the kernel (a
lot of the GDI, for instance, has been separate from the "kernel",
per se, since pre WinNT)
> If they really do create a
>minimal kernel, it perforce will look and behave a lot like the linux
>kernel (or the Solaris kernel, or the Mac OSX kernel, or...). Maybe not
>on the inside, although schedulers are schedulers, maybe not in the way
>it manages processes or memory, but it will need to provide scheduling,
>process management, and memory management, and device management.
It does all this now. The question is whether you can get rid of a
lot of the other stuff, since at a very fundamental level, windows
follows an event driven model, where the events are largely from user
interaction. Inside the guts, a lot of the work is in deciding where
(to which process) to send those events (e.g. a mouse move or click,
>This in turn moves their system closer to being a suitable host for VMs,
>as I think the ideal towards which OS's are moving is the kernel as a VM
>host, everything else as a VM guest on top of a device-independent
>layer. At least I hope we are going there.
I would imagine that this is the case. Certainly, when NT first came
out, there was great emphasis put on the Hardware Abstraction Layer
(so you could run essentially the same OS on Alphas and x86s). The
challenge faced by MS for Windows (as opposed to Linux) is that they
need to have well defined hardware abstraction and the VM to real
hardware layer tightly controlled, because it very important for
digital rights management (to make sure that nobody can put a shim
layer in and tap off the protected content). What is perceived as a
virtue in the Linux world is viewed as a terrifying hole in the MS
Windows DRM world.
More information about the Beowulf