Gail wrote:
> 
> Then, of course, there's Hyper-V built into Server 2008.

Which, according to Hype(tm) is Ultimately New Technology Innovation By 
Microsoft(R), since it uses the Brand-New Never-Before-Seen hypervisor 
technology (which is, still according to the same Hype(tm), pretty much 
similar as in Xen and AFAIK with ESXi, but apparently they don't count, 
since Innovating Sells).

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

Gail wrote:
> 
> Indeed. The problem with multiple SQL instances in VMs is IO. You can 
> allocate CPUs and memory to specific VMs, but there's still a single IO 
> channel, and SQL database tend to be IO bound more often than CPU or 
> memory.

With Xen you can allocate whole disks or RAID-arrays or whatever and 
amount X of RAM to be used straightly by the VM. In my experience the 
lag is mostly unnoticeable (yes, it probably will be more noticeable 
with DB-server, which I'm not running on VMs (well, there are MySQL 
installations on at least 2 VM's, but they're really small DBs), but 
depending on DB-size etc the saved cost of another HW can possibly give 
you even better HW for that one machine to have better IOs for the DB, 
it's not purely black and white).

Then again, MS surely don't want to support Xen, while they're working 
on HyperV - no wonder SQL Server ain't supported on such 
competitor-software ;).

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

"Eero Ahonen" <aer### [at] removethiszbxtnetinvalid> wrote in message 
news:48bc0684$1@news.povray.org...
> Gail wrote:
>>
>> Then, of course, there's Hyper-V built into Server 2008.
>
> Which, according to Hype(tm) is Ultimately New Technology Innovation By 
> Microsoft(R), since it uses the Brand-New Never-Before-Seen hypervisor 
> technology (which is, still according to the same Hype(tm), pretty much 
> similar as in Xen and AFAIK with ESXi, but apparently they don't count, 
> since Innovating Sells).
>

No one ever suggested that MS's marketing dept was 'honest'. I've heard a 
couple of MS developers complaining about the massaged facts the marketing 
dept was pushing.

Post a reply to this message

Darren New wrote:

> How many machine instructions do you think it takes to execute an 
> interpreter for a machine instruction?

How about, hmm, "A LOT"?

The IA32 instruction set is a twisted tangle with no reason or rhyme to 
it. The op-codes follow no particular pattern. The instruction lengths 
very widely. What with MMX, 3D NOW!, SSE, SSE2 and so on, there's dozens 
of registers of all different shapes and sizes. And we haven't even 
talked about all the emulation required for the I/O devices...

> A hundred maybe, but a million?

Last time I checked, the IA32 instruction set is *very* complex. I doubt 
it's got any simpler since then...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>> I've heard of plenty of VMs that run on Linux, but I wasn't aware that 
>> any free ones existed for Windoze.
> 
> Go to microsoft.com and do a search. :-)

Yeah, if I had *wanted* to find one, this would be the first stage. 
However, as I said, I assumed that any VM would be absurdly slow and 
therefore unusable, so I didn't bother looking for one. (Not to mention 
that it's not immediately obvious just how useful these things are until 
you've played with one...)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Darren New wrote:
> Invisible wrote:
>> Presumably that requires *very* specialised hardware though?
> 
> Yes, but nowadays it's built into all the CPUs already. You couldn't do 
> something like this on, say, a 68000 or an 8086.

So when you say "works on any reasonably modern PC", what you *actually* 
mean is "works on any brand new bleeding-edge PC"?

For example, it wouldn't work on the Pentium-III server in my server 
room. (?)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Gail wrote:

> No one ever suggested that MS's marketing dept was 'honest'. I've heard 
> a couple of MS developers complaining about the massaged facts the 
> marketing dept was pushing.

One might argue that if MS were to be "honest" about their products, 
they'd never sell any because they all suck so much...

...especially if that "one" happens to be a Linux advocate. ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Darren New wrote:
> I've been thinking of playing with some virtual machine technology, but 
> I wasn't aware there was one under Windows that would run Linux. Which 
> virtual machine are you running here? (I.e., what's the Windows software?)

	And I'll give another nod to VirtualBox. It's free, and running Windows 
on it was a breeze. I assume the same is the case the other way round.

	Haven't used VMWare products. They're more well known, though.

-- 
Computer Lie #1: You'll never use all that disk space.


                     /\  /\               /\  /
                    /  \/  \ u e e n     /  \/  a w a z
                        >>>>>>mue### [at] nawazorg<<<<<<
                                    anl

Post a reply to this message

"Eero Ahonen" <aer### [at] removethiszbxtnetinvalid> wrote in message 
news:48bc0856$1@news.povray.org...
but
> depending on DB-size etc the saved cost of another HW can possibly give 
> you even better HW for that one machine to have better IOs for the DB, 
> it's not purely black and white).

No, it's not.
In my experience though, the majority of DB servers that have bottlenecks 
have IO ones, even on physical machines. Virtualising those means needing 
separate IO paths, separate physical raid arrays, etc or the virtuals will 
compeate with one another for IO. I know that all the SQL MVPs who have 
given an opinion on this have recommend not virtualising a production DB, 
especially not a large one with lots of activity.
I'd happily virtualise a dev or test environment. Not happy about doing it 
to a heavily-utilised prod server though. Not right now.

> Then again, MS surely don't want to support Xen, while they're working on 
> HyperV - no wonder SQL Server ain't supported on such competitor-software 
> ;).

Well, VMWare's joined the SVVP 
(http://www.windowsservercatalog.com/svvp.aspx?svvppage=svvp.htm), so 
there'll have to be some level of support. How much remains to be seen.
SQL's not supported at all on existing MS virtualisation products (virtual 
server, virtual PC) and it's not currently supported on Hyper-V (which has 
been available for a month or so)

Post a reply to this message

"Invisible" <voi### [at] devnull> wrote in message 
news:48bc09bb$1@news.povray.org...
> Darren New wrote:
>> Invisible wrote:
>>> Presumably that requires *very* specialised hardware though?
>>
>> Yes, but nowadays it's built into all the CPUs already. You couldn't do 
>> something like this on, say, a 68000 or an 8086.
>
> So when you say "works on any reasonably modern PC", what you *actually* 
> mean is "works on any brand new bleeding-edge PC"?

No, he means any modern PC.

> For example, it wouldn't work on the Pentium-III server in my server room. 
> (?)

Sure it would.

Virtual PC requires minimum a 400 MHz Pentium-compatible processor.
VMWare version 5 (previous version) requires a 400MHz processor from the 
following list:
# Intel: Celeron, Pentium II, Pentium III, Pentium 4, Pentium M (including 
computers with CentrinoT mobile technology), XeonT (including "Prestonia")
# AMDT: AthlonT, Athlon MP, Athlon XP, DuronT, OpteronT

Post a reply to this message

"Invisible" <voi### [at] devnull> wrote in message 
news:48bc0a2a$1@news.povray.org...
> Gail wrote:
>
>> No one ever suggested that MS's marketing dept was 'honest'. I've heard a 
>> couple of MS developers complaining about the massaged facts the 
>> marketing dept was pushing.
>
> One might argue that if MS were to be "honest" about their products, 
> they'd never sell any because they all suck so much...

Not all of them.
You're generalising to a hell of a lot of stuff that you've probably never 
used based on your experiences with a couple.

Post a reply to this message

Gail wrote:
> 
> No, it's not.
> In my experience though, the majority of DB servers that have 
> bottlenecks have IO ones, even on physical machines. Virtualising those 
> means needing separate IO paths, separate physical raid arrays, etc or 
> the virtuals will compeate with one another for IO. 

Yep, but think about a small firm, just started. They want reliable 
hardware, they need 2 servers, one for databases and one for something 
else that's not so IO-sensitive (authenticating etc). Now, would they 
get better performance for the DB's by buying 2 individual servers or by 
  using all that money on one server, from which they could utilize most 
of the IO from the DB-engine?

I'd probably go with the latter one. Probably, not surely.

> I know that all the 
> SQL MVPs who have given an opinion on this have recommend not 
> virtualising a production DB, especially not a large one with lots of 
> activity.
> I'd happily virtualise a dev or test environment. Not happy about doing 
> it to a heavily-utilised prod server though. Not right now.

Yep, there's a huge scale on the db-sized used over the globe. Even if 
you'll count out the 2-5 -table one-query-per-day -style ones I 
mentioned earlier.

> Well, VMWare's joined the SVVP 
> (http://www.windowsservercatalog.com/svvp.aspx?svvppage=svvp.htm), so 
> there'll have to be some level of support. How much remains to be seen.

Ahh, nice.

> SQL's not supported at all on existing MS virtualisation products 
> (virtual server, virtual PC) 

As said, AFAIK those are the ones that run over the main OS, they really 
can't pass any unemulated/virtualized command straightly to the hardware.

> and it's not currently supported on Hyper-V 
> (which has been available for a month or so)

Hyper-V is still pretty new and at least I wouldn't be sure if they have 
actually had the time to test it enough. I'd guess the support on 
Hyper-V would get better after some time.

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

Gail wrote:

> 
> No one ever suggested that MS's marketing dept was 'honest'. I've heard 
> a couple of MS developers complaining about the massaged facts the 
> marketing dept was pushing.

Yep.

I haven't used Hyper-V, so I won't say anything about it's quality. It 
might suck, it might be superior, or anything from between. I won't even 
say sure that it's compareable with Xen or ESXi.

It just annoys me while MS continously gets innovations of "new, unseen 
technology" which I'm (or someone else if I know for sure) already using 
and is just unknown for most of the people. Marketing ranting :).

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

"Eero Ahonen" <aer### [at] removethiszbxtnetinvalid> wrote in message 
news:48bc1d26$1@news.povray.org...
>
>  I'd probably go with the latter one. Probably, not surely.

Very likely. Though, as always, test carefully.

> Hyper-V is still pretty new and at least I wouldn't be sure if they have 
> actually had the time to test it enough. I'd guess the support on Hyper-V 
> would get better after some time.

I hope so. There's enough companies fully aboard the vitualisation train 
already, many without checking to see if there's coal,  if the wheels are 
about to fall off or if it's going in the direction they want. <grin>

Post a reply to this message

Gail wrote:
> 
> Very likely. Though, as always, test carefully.
>

Yep.

> I hope so. There's enough companies fully aboard the vitualisation train 
> already, many without checking to see if there's coal,  if the wheels 
> are about to fall off or if it's going in the direction they want. <grin>

As you said: test carefully ;). MS is a big player after all, so before 
they give theier support (which practically means they promise it'll 
work) for such a thing, they'll need to be pretty sure of it. The surer 
they are, the happier will customers be and the better they'll sell.

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

>>  I'd probably go with the latter one. Probably, not surely.
> 
> Very likely. Though, as always, test carefully.

Wait a sec - small firm, just started up, *testing*???

 From what I've seen most *large* companies don't bother to test 
anything, never mind small startups. ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>> One might argue that if MS were to be "honest" about their products, 
>> they'd never sell any because they all suck so much...
> 
> Not all of them.
> You're generalising to a hell of a lot of stuff that you've probably 
> never used based on your experiences with a couple.

Perhaps. But you might argue it's fitting for their most popular 
products (which, presumably, are where they make most of their money).

It wasn't intented to be a serious statement. MS's reputation is the 
stuff of legend - I'm sure we've all been there and done that.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:
> So when you say "works on any reasonably modern PC", what you *actually* 
> mean is "works on any brand new bleeding-edge PC"?

No. I believe it was the 386 that added the capability.

You know, two generations back before the Pentium came out?

Added primarily to emulate multiple "DOS boxes" under Windows.

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Eero Ahonen wrote:
> similar as in Xen 

As far as I understand it, an OS running under Xen has to know it's 
running under Xen, doesn't it?  It's actually quite a large leap from 
"we can run multiple OSes" to "we can run multiple OSes that don't have 
to cooperate with the hypervisor".

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Invisible wrote:
> But surely if you're going to run a guest OS on the physical host CPU,
> the host CPU would need to have hardware support for enforcing the host
> seperation?

Nope. You don't *need* it. You can, for example, get a chunk of code the
guest is trying to execute, translate the little parts that need changing
to provide security and enforcing host separation, then run it. (I don't
even know assembler, so actually I have no idea what I'm talking about)

VMware Inc has dozens of patents on different tricks to do that efficiently.

And yes, there are *also* CPUs (64-bit Intel and AMD CPUs you can buy at any
shop) that include built-in support for virtualization. That just gives you
more performance; doesn't really let you do anything you couldn't do
before...

Post a reply to this message

Invisible wrote:
> Until yesterday, I had always assumed that any kind of emulation or
> virtualisation solution would be roughly 1,000,000x slower than native
> execution, so it's not something I've ever been interested in. Hence
> it's something I know almost nothing about. However, since it appears
> that you can actually run real software at almost native speeds,
> suddenly it becomes far more interesting... ;-)

There is an emulator for PowerPC processors called PearPC.

It has two modes. One is an actual interpreter. Runs anywhere; it's quite
portable C code. 500x slower than native.

The other mode is only for x86 hosts. Translates PowerPC instructions into
x86 instructions on-demand, and caches the translations. 15x slower than
native.

I have also seen an IBM PC emulator written in Java. It was pretty fun to
see Linux and DOS boot inside a browser (even though the RAM usage was
insanely high). But even Lemmings under DOS ran at slow motion in that
emulator...

Post a reply to this message

>> A hundred maybe, but a million?
>
> Last time I checked, the IA32 instruction set is *very* complex. I doubt 
> it's got any simpler since then...

But still, presumably you write your emulator in such a way that the most 
common instructions (that usually take 1 or just a few clock cycles) are 
emulated as fast as possible.

I would guess emulation is a prime candidate to be written in assembler, 
where you can just have a huge branch table for each instruction type, then 
jump the real CPU program counter into that table based on the op-code.  In 
that way, I would imagine you can get away with even a factor of 10 
slow-down for the common instructions.

Post a reply to this message

> 1. Go to http://www1.interq.or.jp/t-takeda/top.html
>
> 2. Download the zip file at the top of that page.
>
> 3. Unzip it somewhere.
>
> 4. Find an ISO image for a Linux LiveCD.
>
> 5. Do the following:
>
>   qemu.exe -cdrom My_LiveCD.iso
>
> 6. Watch it boot up.

qemu doesn't work for me, it crashes out with a "this program has gone 
wrong" windows error box.  The console says "cannot find kqemu" or 
something, so I downloaded and installed that but it still crashes out.  I 
even tried XP compatibility mode etc but nope.  This is on Vista 64bit. 
Shame, because my XP machine is at least 5x slower...

Post a reply to this message

Darren New wrote:
> Invisible wrote:
>> So when you say "works on any reasonably modern PC", what you 
>> *actually* mean is "works on any brand new bleeding-edge PC"?
> 
> No. I believe it was the 386 that added the capability.
> 
> You know, two generations back before the Pentium came out?
> 
> Added primarily to emulate multiple "DOS boxes" under Windows.

The 386 or so added "virtual mode" (or whatever the exact term is) that 
makes the CPU act as if it's in real mode, but it's really in protected 
mode, and the OS can control the page mappings and so forth.

You can use that to make real mode applications (e.g., DOS programs) 
think they have the machine to themselves. But that doesn't work for a 
protected mode program such as an OS.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

scott wrote:

> qemu doesn't work for me, it crashes out with a "this program has gone 
> wrong" windows error box.  The console says "cannot find kqemu" or 
> something, so I downloaded and installed that but it still crashes out.  
> I even tried XP compatibility mode etc but nope.  This is on Vista 
> 64bit. Shame, because my XP machine is at least 5x slower...

On my system I also get an error saying "can't find kemu", but it runs 
anyway.

Try the "-no-kqemu" flag and see if that makes any difference for you...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Nicolas Alvarez wrote:

> I have also seen an IBM PC emulator written in Java. It was pretty fun to
> see Linux and DOS boot inside a browser (even though the RAM usage was
> insanely high). But even Lemmings under DOS ran at slow motion in that
> emulator...

Heh. That's pretty crazy! ;-)

Somebody somewhre is working on a Gameboy emulator in Haskell - but 
AFAIK that machine operates vastly slower than any PC anyway...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

> Somebody somewhre is working on a Gameboy emulator in Haskell - but AFAIK 
> that machine operates vastly slower than any PC anyway...

IIRC it uses a 1MHz processor (can't remember which).  Funnily enough the 
same processor was used in a graphics calculator I once had (TI-86), so it 
was relatively easy to port the games :-)

Post a reply to this message

>>> However, the idea of being able to just plug in a USB stick and click 
>>> an icon and instantly have a complete Linux environment set up 
>>> exactly the way I want it is... pretty neat, actually.
>>
>> You might have problems with IO speed if you do that. USB is a lot 
>> slower than SATA.
> 
> Depends what you want to do. Running Damn Small Linux from a USB drive 
> is probably *faster* than running it from a CD-ROM due to the vastly 
> superior seek time. (CD-ROM drives typically have a horrible seek time!) 
> Also, DSL is based on KNOPPIX and uses a cloop compressed filesystem 
> image, so there's less data to shift (but more CPU load).
> 
> If I get bored enough maybe I'll benchmark it. But either way, it's a 
> trivial matter to copy a folder off the USB drive and onto a spare 
> folder on the local HD, run it for a while, and then just delete it when 
> you're done. That's about the easiest install/uninstall procedure 
> imaginable! ;-)

I was wrong.

DSL copies *the entire CD image* into RAM when it boots. So the medium 
it's running from affects only the time taken to load the whole image 
into RAM; after that, it makes no difference at all.

(In fact, if you copy the CD ISO-image onto your HD and run QEMU, DSL 
boots *faster* than if you really boot from CD.)

In terms of running speed however... Well it turns out that with a few 
mouse clicks you can install Doom under DSL. (The speed with which it 
downloaded and installed shocked me for a moment - until I remembered 
that back when Doom came out, all games came on 1.44 MB floppies, not 
dual-layer BluRay disks like they do today.) Doom runs pretty fast when 
booting DSL natively, and slightly slower under QEMU. (The speed 
difference is small enough that I'm not sure if I imagined it.)

That's a pretty impressove vote for QEMU!

OTOH, my system here at work has very little RAM and a slow CPU, so 
running QEMU basically disables the whole of the rest of the machine. 
You can run QEMU *or* do something else, but not both. (My monster 
machine at home has no such difficulty.)

PS. Yes, I'm playing Doom at work. And? ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

scott wrote:
>> Somebody somewhre is working on a Gameboy emulator in Haskell - but 
>> AFAIK that machine operates vastly slower than any PC anyway...
> 
> IIRC it uses a 1MHz processor (can't remember which).  Funnily enough 
> the same processor was used in a graphics calculator I once had (TI-86), 
> so it was relatively easy to port the games :-)

I'm reasonably sure the Gameboy is a Z80... Wikipedia will know.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:

> I'm reasonably sure the Gameboy is a Z80... Wikipedia will know.

Nope, I was wrong. Apparently it uses a 4 MHz customised Z80 with an 
on-die ROM for bootstrap and a customised instruction set somewhat 
similar to the Intel 8080... Oh well!

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:

> Until yesterday, the only products I'd heard of where Bochs (only runs 
> on Linux) and VMware (AFAIK that's extremely expensive).

I was wrong: Apparently Bochs now runs on Windoze too. (And indeed, QEMU 
is using the Bochs BIOS code, and both use the same VGA BIOS.)

However, Bochs looks significantly more complicated to set up...

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:
>>> Also - virtual machines are SOOOO SLOOOOOW... :-/
>>
>> They aren't that slow. Thing is, you're using QEMU, which isn't a virtual
>> machine, it's an emulator. Software like VMware or VirtualBox use your 
>> real
>> CPU instead of basically executing x86 code on an interpreter.
> 
> Presumably that requires *very* specialised hardware though?

I was right:

http://en.wikipedia.org/wiki/X86_virtualization

Hardware-assisted virtualisation is a feature found only on the very 
newest CPUs from Intel and AMD.

That still leaves open the question of how some products claim to 
virtualise without emulation on CPUs that lack this technology.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

>>> So when you say "works on any reasonably modern PC", what you 
>>> *actually* mean is "works on any brand new bleeding-edge PC"?
>>
>> No. I believe it was the 386 that added the capability.
>>
>> You know, two generations back before the Pentium came out?
>>
>> Added primarily to emulate multiple "DOS boxes" under Windows.
> 
> The 386 or so added "virtual mode" (or whatever the exact term is) that 
> makes the CPU act as if it's in real mode, but it's really in protected 
> mode, and the OS can control the page mappings and so forth.

"Virtual real mode".

http://en.wikipedia.org/wiki/Virtual_8086_mode

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

> Nope, I was wrong. Apparently it uses a 4 MHz customised Z80 with an 
> on-die ROM for bootstrap and a customised instruction set somewhat similar 
> to the Intel 8080... Oh well!

Oh and Wikipedia tells me the TI-86 has a 6 MHz Z80, so I guess not straight 
copy of game-boy apps, some hacks probably needed for timing etc.  Or maybe 
the games just ran a bit 50% faster, I can't remember.

Post a reply to this message

Invisible wrote:
> PS. Yes, I'm playing Doom at work. And? ;-)
> 
S0 you are human after all :-)

John

-- 
"Eppur si muove" - Galileo Galilei

Post a reply to this message

Doctor John wrote:
> Invisible wrote:
>> PS. Yes, I'm playing Doom at work. And? ;-)
>>
> S0 you are human after all :-)

Actually, in truth I was watching the demo. The framerate is a bit too 
low to actually "play" the game. (Plus I don't know the controls.)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:
> Doctor John wrote:
>> Invisible wrote:
>>> PS. Yes, I'm playing Doom at work. And? ;-)
>>>
>> S0 you are human after all :-)
> 
> Actually, in truth I was watching the demo. The framerate is a bit too 
> low to actually "play" the game. (Plus I don't know the controls.)
> 
IIRC the key bindings are configurable. OTTOMH <ESC> in the splash 
screen pulls up a menu and I think options in that menu leads you to key 
bindings. Don't quote me on that though

John

-- 
"Eppur si muove" - Galileo Galilei

Post a reply to this message

"Invisible" <voi### [at] devnull> wrote in message 
news:48bd22a1$1@news.povray.org...
>>
> That still leaves open the question of how some products claim to 
> virtualise without emulation on CPUs that lack this technology.
>

Quote, from the wiki page you linked

"VMware and similar virtualization software for the x86 must employ binary 
translation techniques to trap and virtualize the execution of certain 
instructions. These techniques incur some performance overhead as compared 
to a VM running on a natively virtualizable architecture"
http://en.wikipedia.org/wiki/Binary_translation

Also see
http://en.wikipedia.org/wiki/Operating_system-level_virtualization

The hardware extensions is what makes the hypervisor type products (that run 
'beneath' the OS possible). You'll need that for things like Server 2008's 
hyper-V, VMWare's ESX server (which can make do without), Xen and others.

Things like VMWare workstation can use it but don't need it. MS's Virtual PC 
can't use the hardware emulation at all, though Virtual Server can

Post a reply to this message

Gail wrote:

> Quote, from the wiki page you linked
> 
> "VMware and similar virtualization software for the x86 must employ 
> binary translation techniques to trap and virtualize the execution of 
> certain instructions. These techniques incur some performance overhead 
> as compared to a VM running on a natively virtualizable architecture"

Right. So it only emulates some instructions, instead of all of them?

> The hardware extensions is what makes the hypervisor type products (that 
> run 'beneath' the OS possible). You'll need that for things like Server 
> 2008's hyper-V, VMWare's ESX server (which can make do without), Xen and 
> others.
> 
> Things like VMWare workstation can use it but don't need it. MS's 
> Virtual PC can't use the hardware emulation at all, though Virtual 
> Server can

I've spent some time looking at all the various products out there. It 
seems that QEMU is the only one that can be run without "installing" 
anything. Some install special kernel-mode drivers, others run on the 
bare metal beneigh the first "privilaged" OS, but I'm not seeing 
anything else that you can throw on a USB stick and use anywhere. (I 
also note that several of these things appear to be "based on" QEMU...)

Still, if I was trying to offer hosting of virtual servers, this would 
be a non-issue.

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

Invisible wrote:
> You can use that to make real mode applications (e.g., DOS programs) 
> think they have the machine to themselves. But that doesn't work for a 
> protected mode program such as an OS.

I stand corrected. :-)

-- 
Darren New / San Diego, CA, USA (PST)

Post a reply to this message

Darren New wrote:
> Invisible wrote:
>> You can use that to make real mode applications (e.g., DOS programs) 
>> think they have the machine to themselves. But that doesn't work for a 
>> protected mode program such as an OS.
> 
> I stand corrected. :-)

Some of us have actually *read* the IA32 reference manual. ;-)

[OTOH, I guess some others have written actual IA32 programs, so...]

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message

On Sun, 31 Aug 2008 14:46:44 +0100, Orchid XP v8 wrote:

> OK, so I realise I'm probably about 20 years behind everybody else, but
> seriously... this is THE coolest thing I've seen this year! I would
> never have believed it was physically possible - and certainly not
> without spending hours setting it up. I can't get over how EASY this
> was!

Told ya that Virtualization was useful. :-)

Jim

Post a reply to this message

On Tue, 02 Sep 2008 18:24:52 +0100, Orchid XP v8 wrote:

>> Quote, from the wiki page you linked
>> 
>> "VMware and similar virtualization software for the x86 must employ
>> binary translation techniques to trap and virtualize the execution of
>> certain instructions. These techniques incur some performance overhead
>> as compared to a VM running on a natively virtualizable architecture"
> 
> Right. So it only emulates some instructions, instead of all of them?

Not really; at least as I understand it, if you had to emulate the 
instructions, you'd have to recompile.  What VMware and programs like it 
do is trap them and execute them within their 'sandbox'.  It's not really 
the same as emulation, which tends to be more of a translation process.

Of course, if you have a chip that supports VT (for example), then that's 
entirely different.  Or a dual-core processor/MP system, which I think 
VMware also can take advantage of, by using one core for the virtualized 
system and one for the host.  Not sure how they do that, I don't imagine 
it's well documented publicly.

I do know when I played around with virtualization a few years ago, there 
was a need to recompile kernels that were running (one of the issues that 
prevented NetWare from running natively in XEN, for example - NW kernel 
would've had to be rewritten in order to deal with the fact that it is a 
nonpreemptive OS - from what I understand from people who have worked 
closely with the NW kernel developers, that makes virtualization (at 
least as implemented in XEN 2) more difficult to achieve.

Jim

Post a reply to this message

On Mon, 01 Sep 2008 09:34:34 +0100, Invisible wrote:

> But surely if you're going to run a guest OS on the physical host CPU,
> the host CPU would need to have hardware support for enforcing the host
> seperation?

Nope.  I started running VMware back at version 2.x (now on version 6), 
hardware at that time didn't have any concept of virtualization.  The 
separation was all done in software (and very well isolated as well - 
can't think of any time where the guest took the host down when the host 
was clean).

Jim

Post a reply to this message

On Sun, 31 Aug 2008 12:48:34 -0700, Darren New wrote:

> Orchid XP v8 wrote:
>> OK, so I realise I'm probably about 20 years behind everybody else, but
>> seriously... this is THE coolest thing I've seen this year!
> 
> I've been thinking of playing with some virtual machine technology, but
> I wasn't aware there was one under Windows that would run Linux. Which
> virtual machine are you running here? (I.e., what's the Windows
> software?)

VMware has run Linux under Windows for *ages*.....

Jim

Post a reply to this message

On Mon, 01 Sep 2008 12:57:21 -0700, Darren New wrote:

> As far as I understand it, an OS running under Xen has to know it's
> running under Xen, doesn't it?

I understood that was the case under 2.0, but not sure about XEN 3.x

Jim

Post a reply to this message

Jim Henderson wrote:
> Nope.  I started running VMware back at version 2.x (now on version 6),
> hardware at that time didn't have any concept of virtualization.  The
> separation was all done in software (and very well isolated as well -
> can't think of any time where the guest took the host down when the host
> was clean).

Many security researchers intentionally run malicious and quite dangerous
software inside a VM, so it better have good isolation :)

Post a reply to this message

"Gail" <gail (at) sql in the wild (dot) co [dot] za> wrote:
> "VMware and similar virtualization software for the x86 must employ binary
> translation techniques to trap and virtualize the execution of certain
> instructions. These techniques incur some performance overhead as compared
> to a VM running on a natively virtualizable architecture"

But they're still more than acceptably fast.

Post a reply to this message

Orchid XP v8 wrote:
> 
> Wait a sec - small firm, just started up, *testing*???

Yes. Even them can do thing properly. I know one of those.. Oh, wait, I 
actually own part of it, I was building it ;).

>  From what I've seen most *large* companies don't bother to test 
> anything, never mind small startups. ;-)

Yes. Even them can do things improperly. Luckily the one I work for doesn't.

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message

Darren New wrote:
> 
> As far as I understand it, an OS running under Xen has to know it's 
> running under Xen, doesn't it?  It's actually quite a large leap from 
> "we can run multiple OSes" to "we can run multiple OSes that don't have 
> to cooperate with the hypervisor".
> 

That's the difference, what hardware virtualization support (might be 
that also Xen 3.xx is demanded, as Jim said - dunno, 3.xx has been 
around for long enough for me to gotten started with it) does for Xen. 
Without such, the OS needs to know it's running under Xen, but if you 
have it, you can run unmodified guest OS's also.

-- 
Eero "Aero" Ahonen
    http://www.zbxt.net
       aer### [at] removethiszbxtnetinvalid

Post a reply to this message