[Qemu-devel] International Virtualization Conference

[Qemu-devel] International Virtualization Conference

[Ottavio Caruso]

http://www.linuxworldexpo.de/linux_messe.php?ID=124&STEP=&lang=en

I don't see Qemu mentioned at all. I wonder if any of the developers have been contacted at all. I thing it is a pity that once againg Qemu is ignored.

Ottavio

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Sun, Oct 08, 2006 at 05:30:19AM -0700, Ottavio Caruso wrote:
> http://www.linuxworldexpo.de/linux_messe.php?ID=124&STEP=&lang=en
> 
> I don't see Qemu mentioned at all. I wonder if any of the developers have been contacted at all. I thing it is a pity that once againg Qemu is ignored.
> 
> Ottavio
> 
> 

qemu is primarily a dynamic translator not a virtualizer.

I suppose that qemu could get in by virtue of kqemu. That is closed source but
being closed source hasn't stopped VMware.

> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] International Virtualization Conference

[Joshua Root]

Jim C. Brown wrote:
> On Sun, Oct 08, 2006 at 05:30:19AM -0700, Ottavio Caruso wrote:
>> http://www.linuxworldexpo.de/linux_messe.php?ID=124&STEP=&lang=en
>>
>> I don't see Qemu mentioned at all. I wonder if any of the developers have been contacted at all. I thing it is a pity that once againg Qemu is ignored.
>>
> I suppose that qemu could get in by virtue of kqemu. That is closed source but
> being closed source hasn't stopped VMware.

Or Virtual PC.

The conference looks rather like a marketing-fest. Every one of the
sessions is pushing a particular company's products.

- Josh

Re: [Qemu-devel] International Virtualization Conference

[Jamie Lokier]

Jim C. Brown wrote:
> On Sun, Oct 08, 2006 at 05:30:19AM -0700, Ottavio Caruso wrote:
> > http://www.linuxworldexpo.de/linux_messe.php?ID=124&STEP=&lang=en
> > 
> > I don't see Qemu mentioned at all. I wonder if any of the
> > developers have been contacted at all. I thing it is a pity that once
> > againg Qemu is ignored.
> > 
> > Ottavio
> 
> qemu is primarily a dynamic translator not a virtualizer.

VMware is also a dynamic translator for some code.  "Full"
virtualisation (i.e. not like Xen) requires that.

> I suppose that qemu could get in by virtue of kqemu.

qemu+kqemu is a useful, powerful, versatile x86 virtualisation
package.  It really should be mentioned at a conference on the
subject.  But maybe the conference is only for commercially-backed
parties?

> That [kqemu] is closed source but being closed source hasn't stopped
> VMware.

Don't forget qvm86, which is intended as an open source drop-in
replacement for kqemu.

-- Jamie

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Sun, Oct 08, 2006 at 05:58:51PM +0100, Jamie Lokier wrote:
> Don't forget qvm86, which is intended as an open source drop-in
> replacement for kqemu.
> 
> -- Jamie
> 

One that, when I last checked, was outdated and unmaintained. (Though I
admit I haven't looked at qvm86 recently.)

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] International Virtualization Conference

[Rob Landley]

On Sunday 08 October 2006 10:36 am, Jim C. Brown wrote:
> On Sun, Oct 08, 2006 at 05:30:19AM -0700, Ottavio Caruso wrote:
> > http://www.linuxworldexpo.de/linux_messe.php?ID=124&STEP=&lang=en
> > 
> > I don't see Qemu mentioned at all. I wonder if any of the developers have 
been contacted at all. I thing it is a pity that once againg Qemu is ignored.
> > 
> > Ottavio
> > 
> > 
> 
> qemu is primarily a dynamic translator not a virtualizer.

That's an implementation detail.  The end result is running programs in a 
virtual environment, and qemu's system emulation has lots of virtual hardware 
it attaches to virtual busses, which it performs virtual I/O to, even 
simulating the delivery of virtual interrupts to signal completion of virtual 
DMA.

Rob
-- 
Never bet against the cheap plastic solution.

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Mon, Oct 09, 2006 at 12:05:02AM -0400, Rob Landley wrote:
> > qemu is primarily a dynamic translator not a virtualizer.
> 
> That's an implementation detail.  The end result is running programs in a 
> virtual environment, and qemu's system emulation has lots of virtual hardware 
> it attaches to virtual busses, which it performs virtual I/O to, even 
> simulating the delivery of virtual interrupts to signal completion of virtual 
> DMA.
> 
> Rob
> -- 
> Never bet against the cheap plastic solution.
> 

Here you are using the terms "virtual" and "emulated" interchangably. That's ok
as long as the difference between virtualization and virtual/emulated is understood.

If I follow your logic, then bochs is also a good canidate for the workshop.

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] International Virtualization Conference

[Rob Landley]

On Monday 09 October 2006 8:08 am, Jim C. Brown wrote:
> On Mon, Oct 09, 2006 at 12:05:02AM -0400, Rob Landley wrote:
> > > qemu is primarily a dynamic translator not a virtualizer.
> > 
> > That's an implementation detail.  The end result is running programs in a 
> > virtual environment, and qemu's system emulation has lots of virtual 
hardware 
> > it attaches to virtual busses, which it performs virtual I/O to, even 
> > simulating the delivery of virtual interrupts to signal completion of 
virtual 
> > DMA.
> > 
> > Rob
> > -- 
> > Never bet against the cheap plastic solution.
> > 
> 
> Here you are using the terms "virtual" and "emulated" interchangably. That's
> ok as long as the difference between virtualization and virtual/emulated is
> understood. 

Well, the hardware people see a huge difference.  To them one is "doing it in 
hardware" and the other is "doing it in software".

I stay on the software side, and see them both as different ways to fake an 
execution environment.  Wine fakes a windows system, qemu can fake a 
processor (and then either fake system calls for an app or fake hardware for 
a kernel).

Considering that the original goal of QEMU was "to run the Wine project on 
non-86 architectures" (see 
http://lists.gnu.org/archive/html/tinycc-devel/2003-03/msg00084.html )
I think making a terminology distinction between Wine and QEMU is splitting 
hairs.  (And trying to draw a line between qemu, bochs, and valgrind is 
splitting the result of that.)

> If I follow your logic, then bochs is also a good canidate for the workshop.

If you mean the way Hurd is a candidate for a workshop anywhere Linux is, 
sure.  If it's a purely academic conference where being useful doesn't enter 
into it.  (I followed Bochs and Plex86 5 years ago, but could never actually 
get them to do anything useful despite repeated attempts.  Still haven't, 
although I see Bochs is back from the dead...)

> -- 

And I'm sorry, but I find your tagline actively wrong:

> Infinite complexity begets infinite beauty.

It begets "unmaintainable" after about 5 minutes.

> Infinite precision begets infinite perfection.

You've never been micro-managed, have you?

"Perfection is reached, not when there is no longer anything to add, but when 
there is no longer anything to take away."
 - Antoine de Saint-Exupery

Rob
-- 
"Perfection is reached, not when there is no longer anything to add, but when 
there is no longer anything to take away." - Antoine de Saint-Exupery

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Tue, Oct 10, 2006 at 11:48:33AM -0400, Rob Landley wrote:
> > Here you are using the terms "virtual" and "emulated" interchangably. That's
> > ok as long as the difference between virtualization and virtual/emulated is
> > understood. 
> 
> Well, the hardware people see a huge difference.  To them one is "doing it in 
> hardware" and the other is "doing it in software".
> 

That is not how he uses the terms. He uses them interchangably.

I was just trying to make clear the difference between emulation and virtualization.

> 
> > If I follow your logic, then bochs is also a good canidate for the workshop.
> 
> If you mean the way Hurd is a candidate for a workshop anywhere Linux is, 
> sure.

I was trying to say that qemu (sans kqemu) is a bad candidate. Someone else explains the virtualization-vs-emulation thing much better than I could (short answer: VMware, kqemu, and other virtualizers do it in the hardware whie emulators like qemu and bochs do fully it in the software).

>  If it's a purely academic conference where being useful doesn't enter 
> into it.  (I followed Bochs and Plex86 5 years ago, but could never actually 
> get them to do anything useful despite repeated attempts.  Still haven't, 
> although I see Bochs is back from the dead...)

Someone else pointed it out was more a corporate marketing gig. *shrug*.

> 
> > -- 
> 
> And I'm sorry, but I find your tagline actively wrong:
> 
> > Infinite complexity begets infinite beauty.
> 
> It begets "unmaintainable" after about 5 minutes.
> 

Natural complexity not human-made complexity :)

Think fractals here. Or those pretty pictures we get when looking at subatomic particles.

> > Infinite precision begets infinite perfection.
> 
> You've never been micro-managed, have you?
> 

Really not what I was referring to. :P

> "Perfection is reached, not when there is no longer anything to add, but when 
> there is no longer anything to take away."
>  - Antoine de Saint-Exupery

I agree.

> 
> Rob
> -- 
> "Perfection is reached, not when there is no longer anything to add, but when 
> there is no longer anything to take away." - Antoine de Saint-Exupery
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] International Virtualization Conference

[Rob Landley]

On Tuesday 10 October 2006 1:18 pm, Jim C. Brown wrote:
> On Tue, Oct 10, 2006 at 11:48:33AM -0400, Rob Landley wrote:
> > > Here you are using the terms "virtual" and "emulated" interchangably. 
That's
> > > ok as long as the difference between virtualization and virtual/emulated 
is
> > > understood. 
> > 
> > Well, the hardware people see a huge difference.  To them one is "doing it 
in 
> > hardware" and the other is "doing it in software".
> > 
> 
> That is not how he uses the terms. He uses them interchangably.

A) I'm not a hardware person.

B) The people I've seen care about this are embedded system developers, who 
also make a distinction between "emulator" and "simulator".  (One is a 
hardware board that fakes a certain processor, the other is software that 
does the same thing.  Sometimes, I can even keep them straight.)

> I was just trying to make clear the difference between emulation and
> virtualization. 

I consider this difference an implementation detail that's likely to vanish 
into obscurity as time goes on.

> > 
> > > If I follow your logic, then bochs is also a good canidate for the 
workshop.
> > 
> > If you mean the way Hurd is a candidate for a workshop anywhere Linux is, 
> > sure.
> 
> I was trying to say that qemu (sans kqemu) is a bad candidate. Someone else
> explains the virtualization-vs-emulation thing much better than I could
> (short answer: VMware, kqemu, and other virtualizers do it in the hardware
> whie emulators like qemu and bochs do fully it in the software).   

Modems wandered back and forth between hardware and software before dying.  
Hardware crypto accelerators were really popular a few years back.  One of 
the promises of the cell chip is doing stuff like 3D rending and mp4 
compression entirely in software at a reasonable speed.  And now it's 
only "virtual reality" if you use an actual 3D graphics chip, with software 
rendering it's just "emulated reality".  Right.

The "this must be done in hardware to get reasonable performance" people are 
always amusing, in retrospect.  Personally, I've never bothered to even 
install kqemu.  Maybe when Moore's law stops because we've finally hit atomic 
limits or whatever, I'll start to care.

Rob
-- 
"Perfection is reached, not when there is no longer anything to add, but when 
there is no longer anything to take away." - Antoine de Saint-Exupery

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Tue, Oct 10, 2006 at 10:03:26PM -0400, Rob Landley wrote:
> > That is not how he uses the terms. He uses them interchangably.
> B) The people I've seen care about this are embedded system developers, who 
> also make a distinction between "emulator" and "simulator".  (One is a 
> hardware board that fakes a certain processor, the other is software that 
> does the same thing.  Sometimes, I can even keep them straight.)

They use it differently than we do. See below.

> 
> > I was just trying to make clear the difference between emulation and
> > virtualization. 
> 
> I consider this difference an implementation detail that's likely to vanish 
> into obscurity as time goes on.

The difference between floppy disks and USB flash sticks are likely to vanish
into obscurity as time goes on. Doesn't mean floppy disk designers would get
invited into USB flash sticks conferences (unless they also designed USB
flash sticks).

The way cpu emulation is done in bochs/qemu-softmmu and the way its done in VMware/VirtualPC/etc represent different (though related) software technologies.

The end user doesn't care as long as its fast enough and accurate enough to run
what they want, naturally. Why should they? But developers at a major conference
probably would. They should be able to keep their technology straight.

> 
> Modems wandered back and forth between hardware and software before dying.  
> Hardware crypto accelerators were really popular a few years back.  One of 
> the promises of the cell chip is doing stuff like 3D rending and mp4 
> compression entirely in software at a reasonable speed. 

So?

> And now it's 
> only "virtual reality" if you use an actual 3D graphics chip, with software 
> rendering it's just "emulated reality".  Right.

Virtualization != virtual, so I can't respond to this statement as you wrote it.
I have no idea what "emulated reality" means and I can't see the relationship
between "virtual reality" and "virtualization" other than the fact that both
start with the same 7 letters and the fact that both run on computers.

If I clarify it like this,

: And now it's 
: only "virtualization reality" if you use an actual 3D graphics chip, with software 
: rendering it's just "emulated reality".  Right.

then it makes even less sense.

Quite simply, you are comparing apples to oranges.

Virtualization has nothing to do with virtual/emulated/simulated. Virtualization
refers to running cpu instructions in the guest OS on the host OS's cpu. The
meaning is quite specific here. Do not confuse virtualization with virtual.
They mean two completely different things.

> 
> The "this must be done in hardware to get reasonable performance" people are 
> always amusing, in retrospect.  Personally, I've never bothered to even 
> install kqemu.

Same here. Of course no one is talking about that ... (the fact that qemu
exists is proof that the statement is false imvho).

> 
> Rob
> -- 
> "Perfection is reached, not when there is no longer anything to add, but when 
> there is no longer anything to take away." - Antoine de Saint-Exupery
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.

Re: [Qemu-devel] International Virtualization Conference

[Joshua Root]

Rob Landley wrote:
> On Sunday 08 October 2006 10:36 am, Jim C. Brown wrote:
>> qemu is primarily a dynamic translator not a virtualizer.
> 
> That's an implementation detail.  The end result is running programs in a 
> virtual environment, and qemu's system emulation has lots of virtual hardware 
> it attaches to virtual busses, which it performs virtual I/O to, even 
> simulating the delivery of virtual interrupts to signal completion of virtual 
> DMA.

Part of the generally accepted definition of virtualization is that the
majority of guest instructions execute directly on the real CPU with no
intervention by the VMM. QEMU + qvm86 does count as virtualization if
the system spends most of its time in user mode; QEMU on its own does
not (you run code that is very different to the original binary).

- Josh

Re: [Qemu-devel] International Virtualization Conference

[Rob Landley]

On Tuesday 10 October 2006 5:26 am, Joshua Root wrote:
> Part of the generally accepted definition of virtualization is that the
> majority of guest instructions execute directly on the real CPU with no
> intervention by the VMM. QEMU + qvm86 does count as virtualization if
> the system spends most of its time in user mode; QEMU on its own does
> not (you run code that is very different to the original binary).

So it stops being a virtual environment if you run Java or Python in it?  (or 
anything else that uses bytecode?)

Or if I get one of those old Rockwell Java processors (or a Dallas 
semiconductor Java iButton, or an ARM processor with a J in it) and make a 
coprocessor out of it (I dunno, plug it into the USB port and send code to 
it), I now have a virtual Java environment because the bytecode is running on 
real hardware?

Rob
-- 
"Perfection is reached, not when there is no longer anything to add, but when 
there is no longer anything to take away." - Antoine de Saint-Exupery

Re: [Qemu-devel] International Virtualization Conference

[Jim C. Brown]

On Tue, Oct 10, 2006 at 11:54:57AM -0400, Rob Landley wrote:
> On Tuesday 10 October 2006 5:26 am, Joshua Root wrote:
> > Part of the generally accepted definition of virtualization is that the
> > majority of guest instructions execute directly on the real CPU with no
> > intervention by the VMM. QEMU + qvm86 does count as virtualization if
> > the system spends most of its time in user mode; QEMU on its own does
> > not (you run code that is very different to the original binary).
> 
> So it stops being a virtual environment if you run Java or Python in it?  (or 
> anything else that uses bytecode?)

What do you mean by virtual? Don't confuse virtual with virtualization.

Python and JAVA are not virutalization environments themselves. They don't affect the status of say qvm86 because by the time you get to that level, the bytecodes have already become native machine code anyways.

> 
> Or if I get one of those old Rockwell Java processors (or a Dallas 
> semiconductor Java iButton, or an ARM processor with a J in it) and make a 
> coprocessor out of it (I dunno, plug it into the USB port and send code to 
> it), I now have a virtual Java environment because the bytecode is running on 
> real hardware?

Virtualization refers to running the machine code of a guest OS natively on thoe host OS's cpu. This doesn't apply.

If you ran a JavaOS natively on the coprocessor though, and then ran another JavaOS inside of that, then you get a virtualized Java environment.

> 
> Rob
> -- 
> "Perfection is reached, not when there is no longer anything to add, but when 
> there is no longer anything to take away." - Antoine de Saint-Exupery
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
> 

-- 
Infinite complexity begets infinite beauty.
Infinite precision begets infinite perfection.