Microsoft plans 'hypervisor' for Longhorn

Microsoft plans 'hypervisor' for Longhorn

[Christopher S. Aker]

http://news.com.com/2061-10794_3-5735471.html

Article mentions Xen as Microsoft's competition.

-Chris

Re: Microsoft plans 'hypervisor' for Longhorn

[M.A. Williamson]

> Article mentions Xen as Microsoft's competition.

The MS hypervisor is a very interesting beast from our PoV: * drivers run 
in a dom0-like "parent" partition (cut down Longhorn) * "child" partitions 
devices are plumbed through this * VMM-aware OSes can use APIs called 
"enlightenments" (don't you usually achieve enlightenment through 
practising Zen? :-D) to avoid full-virt penalties

Their roadmap includes live migration and suspend-resume, USB 
virtualisation, etc.

It'll be interesting to see if they require hypervisor-enabled hardware or 
if they pull in the binary scanning / rewriting from VirtualPC. It'll also 
be interesting to see just how open their "open" device / VMM interfaces 
are: in particular, can Linux support them as a guest, can Xen fake them 
out to boost performance when running Longhorn guests?

Things are evidently heating up in virtualisation land. It'll be 
interesting to see what VMWare's response to this is.

Cheers,
Mark

Re: Microsoft plans 'hypervisor' for Longhorn

[Jacob Gorm Hansen]

M.A. Williamson wrote:
>> Article mentions Xen as Microsoft's competition.
> 
> 
> The MS hypervisor is a very interesting beast from our PoV: * drivers
> run in a dom0-like "parent" partition (cut down Longhorn) * "child"
> partitions devices are plumbed through this * VMM-aware OSes can use
> APIs called "enlightenments" (don't you usually achieve enlightenment
> through practising Zen? :-D) to avoid full-virt penalties
> 
> Their roadmap includes live migration and suspend-resume, USB
> virtualisation, etc.

I like live migration :-) . Do you have a link to this roadmap?

Jacob

Re: Microsoft plans 'hypervisor' for Longhorn

[M.A. Williamson]

> I like live migration :-) .

Yes, I suppose you do :-)

> Do you have a link to this roadmap?

Sure thing!

http://download.microsoft.com/download/9/8/f/98f3fe47-dfc3-4e74-92a3-088782200fe7/TWAR05013_WinHEC05.ppt

Sorry it's so long. If you go through a sane route of their site you might 
find it more at a more straightforward URL. Suffice to say, it's a 
presentation on the hypervisor from WinHEC, which they appear to have made 
public with all this (sudden) publicity.

All the juicy stuff is in there :-)

Cheers,
Mark

> Jacob
> 

Re: Microsoft plans 'hypervisor' for Longhorn

[M.A. Williamson]

Replying to myself again...

> It'll be interesting to see if they require hypervisor-enabled hardware 
> or if they pull in the binary scanning / rewriting from VirtualPC.

Ah, it seems they will be requiring Vanderpool / Pacifica (although by 2007 
this'll be less of a restriction).

Cheers,
Mark

Re: Microsoft plans 'hypervisor' for Longhorn

[Jon Smirl]

On 08 Jun 2005 02:21:04 +0100, M.A. Williamson <maw48@cam.ac.uk> wrote:
> > Article mentions Xen as Microsoft's competition.

Could Apple's x86 switch require Intel VT too? The one year delay
lines up with VT availability.  Requiring Intel VT would probably lock
out all existing machines since it would be difficult to emulate on
existing chips. Intel VT specific shuts out AMD which matches their
announcement.

The simple reason for requiring VT is to get MS Windows app support.
Windows would run in another VM. A virtual graphics card would then
make it appear inside the Mac display. Running an SMB server
internally allows partitions to be shared. Daemons in each OS could
export the clipboard over a virtual network to make DND work and do
things like mouse handoff.

Any rumors of Apple working on a hypervisor?

-- 
Jon Smirl
jonsmirl@gmail.com

Re: Microsoft plans 'hypervisor' for Longhorn

[M.A. Williamson]

> Could Apple's x86 switch require Intel VT too? The one year delay
> lines up with VT availability.  Requiring Intel VT would probably lock
> out all existing machines since it would be difficult to emulate on
> existing chips. Intel VT specific shuts out AMD which matches their
> announcement.

I guess... But OTOH, it'll probably take that long to get the 3rd party 
developers up to speed, develop and launch new machines, etc. so it could 
be coincidence I suppose.

> The simple reason for requiring VT is to get MS Windows app support.
> Windows would run in another VM. A virtual graphics card would then
> make it appear inside the Mac display. Running an SMB server
> internally allows partitions to be shared. Daemons in each OS could
> export the clipboard over a virtual network to make DND work and do
> things like mouse handoff.

The ability to do that sort of thing is (IMHO) one of the big wins of 
moving to x86, as far as Apple is concerned.

> Any rumors of Apple working on a hypervisor?

I haven't heard any. It'd make good sense, though. OTOH, I guess they might 
have talked to VMWare - the VMWare workstation product would also be a good 
fit.

Of course if they do use a hypervisor they should really use Xen ;-) Who 
knows, perhaps they're secretly working on this - stranger things have 
happened.

Cheers,
Mark

Re: Microsoft plans 'hypervisor' for Longhorn

[Jacob Gorm Hansen]

Jon Smirl wrote:
> On 08 Jun 2005 02:21:04 +0100, M.A. Williamson <maw48@cam.ac.uk> wrote:
> 
>>>Article mentions Xen as Microsoft's competition.
> 
> 
> Could Apple's x86 switch require Intel VT too? The one year delay
> lines up with VT availability.  Requiring Intel VT would probably lock
> out all existing machines since it would be difficult to emulate on
> existing chips. Intel VT specific shuts out AMD which matches their
> announcement.

I had a look at some of their developer docs and they were IA-32
specific. I don't think Apple has a VMM plan at the moment, and I don't
think they care too much about running Windows apps as they would fear
that the presence of good Windows emulation would reduce the number of
native apps for their system. I think they will do everything they can
to keep their machines special, to keep their brand value.

I would not buy Apple stock at this point, their plan seems extremely risky.

Jacob

Re: Microsoft plans 'hypervisor' for Longhorn

[Mark Williamson]

> I had a look at some of their developer docs and they were IA-32
> specific. I don't think Apple has a VMM plan at the moment, and I don't
> think they care too much about running Windows apps as they would fear
> that the presence of good Windows emulation would reduce the number of
> native apps for their system.

Yes, I agree - I guess they might well avoid VMMs for that reason, at least 
for the time being.  I wonder if perhaps VMWare is considering a port though 
- it'd certainly make some sense for them now.

> I think they will do everything they can 
> to keep their machines special, to keep their brand value.
>
> I would not buy Apple stock at this point, their plan seems extremely
> risky.

On the one hand, from the consumer PoV their new machines will not look much 
different: cool looking, friendly machines that won't run Windows software.  
OTOH, the PR issues and the logistical (and marketing) problems with the 
actual transition could be *really* painful.  We live in interesting times...

Cheers,
Mark

Re: Microsoft plans 'hypervisor' for Longhorn

[Harry Butterworth]

On Wed, 2005-06-08 at 13:27 +0100, Mark Williamson wrote:

> OTOH, the PR issues and the logistical (and marketing) problems with the 
> actual transition could be *really* painful.  We live in interesting times...

Apple already managed a transition from 68000 to PPC.  If they did it
before, they can do it again.  Probably easier.

Re: Microsoft plans 'hypervisor' for Longhorn

[Jacob Gorm Hansen]

Harry Butterworth wrote:
> On Wed, 2005-06-08 at 13:27 +0100, Mark Williamson wrote:
> 
> 
>>OTOH, the PR issues and the logistical (and marketing) problems with the 
>>actual transition could be *really* painful.  We live in interesting times...
> 
> 
> Apple already managed a transition from 68000 to PPC.  If they did it
> before, they can do it again.  Probably easier.

Their marketshare was a lot bigger back then. While I am sure a lot of
people have just started to consider buying their first Mac, due to
spyware and worm problems with Windows, it seems to be a risky thing to
do right now, as you know your machine will be EOL'ed in a year.

Jacob

Re: Microsoft plans 'hypervisor' for Longhorn

[aq]

On 08 Jun 2005 02:21:04 +0100, M.A. Williamson <maw48@cam.ac.uk> wrote:
> > Article mentions Xen as Microsoft's competition.
> 
> The MS hypervisor is a very interesting beast from our PoV: * drivers run
> in a dom0-like "parent" partition (cut down Longhorn) * "child" partitions
> devices are plumbed through this * VMM-aware OSes can use APIs called
> "enlightenments" (don't you usually achieve enlightenment through
> practising Zen? :-D) to avoid full-virt penalties
> 
> Their roadmap includes live migration and suspend-resume, USB
> virtualisation, etc.
> 
> It'll be interesting to see if they require hypervisor-enabled hardware or
> if they pull in the binary scanning / rewriting from VirtualPC. It'll also
> be interesting to see just how open their "open" device / VMM interfaces
> are: in particular, can Linux support them as a guest, can Xen fake them
> out to boost performance when running Longhorn guests?

yeah, i like this point: the chance for Xen to fake their hypervisor
to run Longhorn is really interesting ;-)

regards,
aq