Talk:Comparison of platform virtualization software

USB support of qemu

I don't think there should be a "yes" for qemu. Just take some usb printer and press on "print" and you'll get a bluescreen under WindowsXP as guest OS. —Preceding unsigned comment added by 62.134.177.134 (talk • contribs) 15:23, August 1, 2007

OOPS so sorry...

i didn't read this page before adding colinux and user-mode linux

but i'll keep them on the main page...you can delet them if you want

we don't need to vote here this is a non-sens what we need is a clear definition of virtual machine...

mabe the fronteer is tigh between all the notions...

users are going on this page in order to choose a virtual machine ornvidia search information about...see comparaison of operating system or filesystem for reference...

so ,my idea is :

• changing the name of the page into a more generalist topic in order to include things such wine,SFU...
• do a clasification of theses "virtual machine" with the folowing criter:
  • Possible Use (for example bosh can serve as a machine emulator...in order to port an os to an architecture ...wine can't do that)
  • technology (machine emulator,wrapper(wine)...)

theses two categories will be merged into only one

in order to do this clasification i propose the folowing presentation: I HAVE A PROBLEM WITH TABS...PLEASE LOOK AT THE CODE IN ORDER TO KNOW WHANT I WANT TO TELL

(optional comparaison of the different general technologies(wrapper,virtual machine) showing what is diferent between all them)

MAchine emulator

(optional description of what is a "machine emulator")

Name	Creator	Host Processor	Guest Processor	Host OS	Officially supported guest OS	Guest OS SMP available?	Runs Arbitrary OS	Drivers for supported guest OS available?	Method of operation	License	Typical use	Guest OS speed relative to Host OS
Bochs	Kevin Lawton	x86, x86_64, Sparc, PowerPC, Alpha, MIPS	x86, x86_64	Windows, Linux, OS X, IRIX, AIX, BeOS	DOS, Windows, xBSD, Linux	Yes	Yes	?	Emulation	LGPL	Hobbyist, Developer	Very slow

Wrapper

(optional description of what is a wrapper)

Name	Creator	Host Processor	Guest Processor	Host OS	Officially supported guest OS	Guest OS SMP available?	Runs Arbitrary OS	Drivers for supported guest OS available?	Method of operation	License	Typical use	Guest OS speed relative to Host OS
wine	???	x86	x86	Linux,unix	Windows(win32 subsystem only)	???	???	?	wrapper	???	run win32 apps under linux/unixes	??? insert a % here

And I propose to migrate this "Expert debat" to the Virtual machine page

But in order to better understand what a virtual machine is we need example that are present on the "Virtual machine page" but we could,in order to show well the diferences between the term, add a column and put inside "virtual machine" "wrapper",because technology is more precise and give infotrmation about the exact system that use the "virtual machine" and the use is too general

Another idea is explainig the difference under the title ""==wrapper==" or "==virtual machine== —Preceding unsigned comment added by 81.241.20.196 (talk • contribs) 22:25, January 7, 2006

Would also suggest that a column to indicate whether vm's are portable to other physical machines would be useful, as some would feel that this is a major decision factor (i.e. never having to go through the pain of recreating all your virtual machines when upgrading hardware, etc). Several years ago, this would have been a deciding factor in favor of VMware (fully portability) vs Virtual PC (limited portability, same processor type only). Situation may have changed several versions on, but you get the idea. —Preceding unsigned comment added by 124.168.215.162 (talk) 22:10, 28 April 2008 (UTC)

Faster than native?

The page says that the speed of UML is "Native** (some people says that it is faster than natively)". How's that possible, faster than native? IMHO UML is "Slow". —Preceding unsigned comment added by 82.192.228.162 (talk • contribs) 13:47, January 18, 2006

Virtual PC vs VMWare

I notice that they both say "close to native" -- I was doing a quick bit of research here before deciding which one to use, and based on that I took Virtual PC because I have access to it through MSDN. Big mistake. I've just abandoned Virtual PC for VMWare because there is no way Virtual PC can be described as close to native (yes, I installed the extensions). It runs at a reasonable speed, but installing Windows, Visual Studio etc. takes HOURS, whereas in VMWare it was, well, close to native. —Preceding unsigned comment added by 218.101.44.115 (talk • contribs) 22:51, February 1, 2006

The reason is that VirtualPC doesn't run very fast is that it only virtualises rings 1-3 of the x86 architecture. VMWare does "pseudo" virtualisation of ring 0 too (the guest operating system's core kernel). Virtual PC emulates ring 0 making installation of the operating system, and disk heavy operations like installation of big packages very slow.

qemu with kqemu does dynamic translation of ring 0 code making it slightly faster. More recent versions of qemu also do "pseudo" virtualisation of ring 0 when qemu is run with the -kernel-kqemu option making it run much faster (except that the peripheral emulation is still a little slow). This should be added as another qemu row (qemu w/ kqemu module and -kernel-kqemu option) where it runs at "near native" speed. —Preceding unsigned comment added by 86.130.221.127 (talk • contribs) 19:51, June 12, 2006

I have to disagree with this, I tried VMWare 5.5 and 6 today with Windows XP as a guest (I have XP as dual boot as well for speed comparison) and it was nothing like Native. The window rendering being slow didn't help either (yes, VMWare Tools were installed). Enverex 23:32, 31 December 2006 (UTC)

Added notes on 3/23/2007 - I concur with above. VMware can be near native performance, but very frequently is not, especially in workloads with substantial context switching or I/O. Same with Xen. FWIW, mainframe z/VM row didn't have NPOV, and exaggerated the performance characteristics. jsavit. —Preceding unsigned comment added by Jsavit (talk • contribs) 20:47, March 23, 2007

Additional sub-category needed

It appears to me that as currently formulated, the page makes no mention of virtual machines that don't purport to emulate the instruction set of any physical CPU, but are created specifically to be virtual machines. Examples include the Java VM and the UCSD p-machine. Given that the title of the page remains "Comparison of virtual machines," clearly these also qualify. User:Dfavro 12:39, 1 April 2006 (UTC)

There seem to be three conceptually different things lumped together here; emulators (BOCHS, etc), hypervisors (Xen, VMWare, etc), and extended compartmentalisation facilities in an OS (FreeBSD Jails, Solaris Zones, etc). I would recommend splitting this article into three tables, or possibly splitting it into three new articles and turning this one into comparison of the three different methods. --David Chisnall 18:36, 8 March 2007 (UTC)

I second that. It's very hard to compare feature between similar "virtualisation" applications because the table mixes these different things mentioned above. I'm not even sure if it's correct to use the term "vitrualisation" for emulators - I always seen them as something different (although a few emulators allow running in a virtualized way when host and target are the same kind of hardware - this is usually an add-on feature though and they should be seen as emulators with visualization capabilities).

Similarly, OS containers are a different thing - applications still runs on the same system, but resources are allocated and external things are hidden. This can possibly fit within the "virtualisation" definition but is still very different than virtual machines since they can't run any real-mode code.

216.94.210.146 (talk) 14:49, 2 April 2009 (UTC)

I agree with these ideas. Breaking it down to groups that are actually similar is important. I also believe this can be done in one article to make it easier for the person learning about the subject. - KitchM (talk) 22:08, 8 April 2011 (UTC)

Subjective use of green and red

Within the table, some VM qualities are marked with red or green backgrounds, presumably to indicate that these traits are desirable or undesirable in an emulator. These appear to be set with the assumption that the user is looking for the "best" emulator, as though this were a product comparison. It should be more objective than that. If a reader is looking for a specific type of emulator for a specific project, some "undesirable" qualities may in fact be desirable, or meaningless, depending on the application. I would like to propose the removal of subjective indicators within the VM comparison. "Just the facts, ma'am." Fastolfe00 17:20, 17 April 2006 (UTC)

Proposed move

The page's name is nice and short but wouldn't the name Comparison of various virtualization and emulation tools be more descriptive? --unforgettableid | talk to me 04:53, 27 April 2006 (UTC)

Agree. "Virtual machine" sounds a bit misleading.

The only thing is I'd remove the word "various" from the title as it does not adds value. --K001 21:00, 5 July 2006 (UTC)

(about the content of the article) SSE-2 in the table?

should SSE-2 support be in the comparison table?

--DDDW 21:23, 26 May 2006 (UTC)

UML can't be considered as fast as OpenVZ or Linux-Vserver

As the subject says.

UML boots a whole nother kernel and runs things usually from a loop device.

OpenVZ and Linux-Vserver both use the host kernel, and the host filesystem, employing a more "chroot" approach.

Just my tuppence. —Preceding unsigned comment added by 193.49.124.107 (talk • contribs) 15:26, June 7, 2006

Missing information on the early VM systems

Historical information, to the best of my recollection and research --

Name	Creator	Host Processor	Guest Processor	Host OS	Officially supported guest OS	Guest OS SMP available?	Runs Arbitrary OS	Drivers for supported guest OS available?	Method of operation	License	Typical use	Guest OS speed relative to Host OS
CP-40	R. J. Adair, R. U. Bayles, L. W. Comeau, R. J. Creasy (IBM CSC)	IBM System/360-40 (modified)	System/360 Basic	CP-40	CMS, OS/360, DOS/360	No	Yes	(none required)	Virtual Machine Hypervisor	Copyright Only	Research, Developer	up to 95%
CP-67/CMS	Adair, Bayles, Comeau, Creasy, et al (IBM CSC, MIT Lincoln Lab)	IBM System/360-67	System/360 Basic	CP-67	CMS, OS/360, DOS/360, CPREMOTE	Experimental	Yes	(none required)	Virtual Machine Hypervisor	Copyright Only	Research, Developer, Production	up to 95%
VM/370	S. R. Newson, R. A. Seymour, C. Young, et al (IBM)	IBM System/370-VS (135/145/155-II/165-II/158/168)	System/370 Basic, System/370-VS	VM-CP	CMS, RSCS, VMREMOTE, any S/370 OS including itself	No	Yes	(none required)	Virtual Machine Hypervisor	IBM SCP, Copyright	Production, Research, Developer	up to 95%

Dave Tuttle 19:42, 13 October 2006 (UTC)

Dave Tutle knows whereof he speaks, as one of the original developers of CP/67 and VM/370. I would just quibble that the "up to 95%" lacks the balancing "as low as..." which could be experienced, especially with MVS guests before enhancements like shadow page tables. jsavit, 3/23/2007 —Preceding unsigned comment added by Jsavit (talk • contribs) 20:51, March 23, 2007

The earliest example I can think of was the IBM 1410 able to run in what was called by IBM "1401 emulation mode." These two machines were similar 6 bit character based machines with different address lengths.

Also, the SDS 940 and the DEC PDP-10 and their respective operating systems for the early time-share industry preceded the IBM 360/67 and the CP-67 OS in support for paging registers and address-space virtualization if I recall correctly. Halwyman (talk) 18:02, 24 December 2007 (UTC)

VM/370 an open source operating system

VM/370 is an open source virtual machine operating system. While for commercial purposes, zVM replaces VM/370 (as well as VM/BSE, VM/HPO, VM/XA, and VM/ESA) the latter is available and used (even on x86 and x86-64 hardware via an emulator) by individuals.

Please note on the above section, VM/XA and VM/ESA also ran on System/390 hardware before the release of zVM (this is not listed). Also note subsequent versions of VM (VM/XA and VM/ESA) allowed for defining virtual processors and hence guest operating systems to run in SMP mode.

--Bmoshier (talk) 11:48, 1 January 2008 (UTC)

Table header hard to read

Since the table is so long, I like that the header row is repeated for readability. However, I missed this when I first read through the table because the header rows don't stand out against the data in the table well. How about we color-coat the header rows with a light-gray or similar color to clearly distinguish the column labels from the table data? --Pekster 04:14, 7 January 2007 (UTC)

Readability is horrible - would it be better if X/Y axes were swapped? 203.113.233.59 22:47, 23 March 2007 (UTC)

Differences between the various aproaches to virtualisations

The article does not realy explain the differences between the different virtualisations. Also, when I read that nearly every vitrtualisation is working at "native speed" or "near native speed", this makes the comparison rather useless and looks to me as if every product vendor has updated the material to make his solution look good so much, that the comparison looks rather useless. Also, what is the use to have x versions of VMware products in the comparison, when even basic products are missing or one could at least as useful also distinguish betwen XEN 2 and XEN3 or between Solaris Zones and Partitions?

As for the basic information, I think the article should explain that basicaly, there is a trade off between flexibility (by emulating more) and speed (which is gained by emulating less).

• So naturaly, solutions that even emulate the Processor are slowest.
• Next come VMs that emulate the hardware except the CPU (MS Virtual Server and VMware, where the MS product is still much slower than VMware).
• Then comes XEN with its paravirtualisation. It doesn't emulate the hardware but allows different kernels per VM. However, at present (in future it will also support Windows via software only on Novell Suse do to an agreement between Novell and MS) it only virtualises Windows on Chips supporting it, otherwise it only serves Linux/Solaris Guests.
• And speedwise best are those solutions that do the virtualisation at OS level, as they have the least to simulate (however that also means that you can't have different OS/different kernels on one machine).

So for development and testing, as well as for a desktop machine where you just want to use some software form another OS or when you want to consolidate machines with different OS on one machines, the solutions with less speed and more options are better. This (especialy Desktop and testing) is also the world, where most virtualisation aproaches which are today popular in the Intel world come from (VMware, Virtual PC, Parallels etc).

However, when you host lots of standard enviroments (hoster, but also standardized enterprise server), you should take a closer look toward the OS virtualisation solutions.

194.138.39.53 15:25, 15 March 2007 (UTC) Kai Froeb http://kai.froeb.net

I agree that speed comparisons are incorrectly stated. I would rather see a scale that assumes nothing is as fast as without virtulization. (Therefore, there can be no so-called "native" speed relationship. The question always comes up, "Native to what?".) With that standard, the scale should note how many relative steps below normal it takes the speed of processing. Perhaps a simple 1 to 10 or just 1 to 5 scale would be sufficient. - KitchM (talk) 22:19, 8 April 2011 (UTC)

Intel x64 emulation limitations

VMware technical FAQ states that running 64-bit guests in 64-bit hosts environment requires VT support. That means, cheap Core 2 Duo E4300/E4400 CPU's won't be able to run 64-bit guests at all (due to lack of proper segmentation in 64-bit mode). Is there any emulator which allows to run 64 bit guests on such CPU's with reasonable speed? Can QEMU or VirtualBox do that? QuestPC 03:22, 31 May 2007 (UTC)

Qemu can emulate a 64-Bit CPU on a 32-Bit CPU, speed OTOH - no. It's theoretically possible, VMware does it on the 32-Bit arch, but I doubt anyone are willing to invest the effort when VT is readily available. VT can even run 64-Bit guests on 32-Bit hosts.--Anss123 05:06, 31 May 2007 (UTC)

No, of course I understand that emulating 64-bit CPU using physical 32-bit CPU is a bad idea. I am more concerned about new and future low-cost Intel 64-bit CPU's, which wouldn't have VT enabled. There will be more such CPU's besides Allendaie, single core Conroe-L, and so on. How much slow would they run QEMU or VirtualBox 64bit on 64bit without VT? You believe there's no hope? QuestPC 13:40, 31 May 2007 (UTC)

VT does not, at this point in time, offer any increase in performance over VMware/KQEMU or similar solutions. What it does offer is a second interrupt table, which makes implementing a virtualizer a comparable ease contra entirely software based approaches. In fact, the reason you need VT to virualize a 64-Bit CPU is because no one have gone through the trouble of implementing a software based vitualizer for 64-Bit CPUs (and I doubt anyone ever will).

If you are satisfied emulating a 32-Bit CPU, VMware will happily run on a E4300 - full speed (~90% of native).

QEMU can emulate many different forms of CPUs, at about ~10 percent native speed. A VT enabled CPU does not offer anything useful to QEMU, so E4300's lack of VT is irrelevant. This is because QEMU is an emulator, not a virtualizer.

I've never used VirtualBox, but I assume it's one of the many virtualizers that popped up after VT hit the streets. In that case, VirtualBox will not run at all on a E4300.

--Anss123 14:46, 31 May 2007 (UTC)

No, I am not satisfied emulating only a 32-bit CPU with real 64-bit CPU. My really old P3 Tualatin is enough to emulate 32bit. Also, 64-bit Linux is becoming more widespread, and future version of Windows expected to come only 64-bit. So, there's real need to emulate 64 on 64. KQEMU performance on my old CPU is reasonably good. I wonder whether it's possible to use 64-bit KQEMU on E4300 to achieve similar level of performance on E4300, without VT. And yes, I know that VMware works slower with VT, but that's not matter - anyway it doesn't run 64-bit guests with X64 without VT (like E4300). You can read about VirtualBox here. QuestPC 03:47, 1 June 2007 (UTC)

I see VirtualBox is a variant of KQEMU, which makes it a 32-Bit virtualizer. As I said earlier, no one have gone through the trouble of implementing a software based vitualizer for 64-Bit CPUs and I doubt anyone ever will. So no, KQEMU/VirtualBox can not and will not do 64-Bit code on a E4300.

--Anss123 06:31, 1 June 2007 (UTC)

Reliable sources for factual claims?

Please see Wikipedia talk: WikiProject Software where I have started a discussion on reliable sources in the context of software articles and software comparisons. I am pointing this out because I criticised this software comparison for making unsourced, contentious claims about performance.—greenrd 12:27, 7 June 2007 (UTC)

As far as I can tell the performance numbers are made up like this:

Paravirtualization: Native

Virtual machine: Near native

Dynamic recompilation: ~10%

Interpretive emulation: Slow

An article comparing the various methods of emulation might be of more value than the unsourced performance figures. On a different note, the term Virtual Machine generally don't extend to emulators such as DosBox, QEMU and Botch. If this article is to include emulators, it currently leaves out UAE, Shapeshifter and countless others. Might as well cut them out, and skirt around the performance figure issue.

--Anss123 16:15, 7 June 2007 (UTC)

Cleanup request

Just regarding the table itself: 1. it's too wide, 2. cells seems to be missing. Said: Rursus ☺ ★ 14:26, 8 June 2007 (UTC)

I think it is cleaned up enough now, so I am removing the cleanup request. CompotatoJ (talk) 19:46, 18 May 2008 (UTC)

Can we please switch to HTML table markup? I just edited the table and I am disgusted as from how complicated it is. Much easier would be if we see the columns already in the markup like this:

blah

blubb

--ThorstenStaerk (talk) 07:42, 10 April 2009 (UTC)

3D Acceleration

I think there should be a column for 3D Acceleration support, as it seems it is a rather rare feature of Virtual Machines. —Preceding unsigned comment added by 59.101.34.216 (talk • contribs) 05:18, June 21, 2007

Merge proposal

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section.

Recently a new page Comparison of virtual machines features was created, that describes some additional features of virtual machine software. I suggest this is merged into this article. We could make it into two tables: one with general information, and one with features such as "Runs Arbitrary OS" and some of the features from the new article. – Chip Zero 15:33, 2 July 2007 (UTC)

Support. Comparison of virtual machines features is quite sparse at the moment. Additionally, its intentions and content are very similar to comparison of virtual machines and could be integrated quite easily. Merge. Freedomlinux 02:59, 11 July 2007 (UTC)

Support. I originally created this feature comparison as new columns to the table and they got deleted with comment "it's an additional feature, not a base tech info". I agree there were already many columns and I did not thought of a second table, so I created a new page. I find this feature comparison important when choosing a Virtual Machine, but I have no preference between "two pages" and "two tables". Jrouquie 12:27, 16 July 2007 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

I merged the table into a new section. Someone may wont to look into moving some columns from the "general" table to other table, but at least all the information is on one page now. – Chip Zero 13:55, 21 July 2007 (UTC)

Removed GUSS

It appears to have been abandoned before it was at all usable... the list ...

 http://lists.gnu.org/archive/html/guss-hackers/

... seems to be almost exclusively spam for the last half decade. So I figured it didnt have sufficient Notability. --Treekids 18:47, 23 July 2007 (UTC)

"Runs Arbitrary OS" vs paravirtualization

Several systems say they do paravirtualization, yet claim they support arbitrary guests. That seems to be contrary to the nature of paravirtualization. —Preceding unsigned comment added by 72.254.27.202 (talk • contribs) 0:55, 7 August 2007

Indeed, for paravirtualization, they require OS support. But if an OS does not support paravirtualization, they may be able to fall back to regular virtualization. This is at least the case for VMWare, but I'm not sure about all the other systems that are said to do this. – Chip Zero 09:09, 7 August 2007 (UTC)

Error in table : VMware Player /can/ "boot an OS on another disk partition as guest"

Why does the last comparison table say No (red background) for "Can boot an OS on another disk partition as guest" regarding VMware Player ? I'm typing this a VM run by player, host is Linux running from one partition of an IDE disk and guest is Windows 2000 Pro installed on another partition of the same disk (aka "raw" disk). Yes, such settings can be tricky (and largely rewarding), for sure the Player runs them as well as Workstation. I'm not editing the table myself, for fear of messing with the layout; I think someone should correct that error anyway. -- 90.31.227.59 22:15, 10 September 2007 (UTC) Ninho

FreeBSD Jail

Hey, can someone explain me why jail is a "virtual machine" ?

I am FreeBSD user, and i am thinking it cannot be compared, because is only process/es separation facility not emulation of new kernel "instance".

It is chroot+additional_restrictions.

It should be deleted from this table, if not why "chroot" is not there ? —Preceding unsigned comment added by 88.12.214.196 (talk) 23:00, 12 September 2007 (UTC)

Good point. VMs are often used as 'jails' and I can see how one can mix them up, especially if one add paravirtualization to said mix. Not even sure if I'm able to keep them separate in my own head.

I say remove jail, it's more like a restricted application than a virtual machine.

--Anss123 11:04, 1 October 2007 (UTC)

If FreeBSD jail is not going to be removed, then the supported guest architecture entry should at least be changed to indicate that the "guest" operates under the same ISA as the host. — Preceding unsigned comment added by 76.226.239.235 (talk) 05:11, 21 July 2011 (UTC)

Live Migration

It would be nice if live migration capability of a VM was in the table. I know that XEN, OpenVZ and VMWare ESX are capable of live migration. —Preceding unsigned comment added by 194.78.219.251 (talk) 09:58, 1 October 2007 (UTC)

What does "Runs Arbitrary OS" mean?

This column seems to have no meaning. Of the 30 rows that specify "Yes," there seems to be a wide range of interpretation. One of them even has a "Yes" but lists only one OS that it can run!

If this column has meaning, it should be explained on the page somewhere so that it can be interpreted consistentlyby readers and editors. If it doesn't have any meaning, it should be removed. —Preceding unsigned comment added by 192.18.128.5 (talk) 18:10, 19 October 2007 (UTC)

My aim in putting that column in when I created this page was to have some way of showing that the VM in question could run an OS that its developers had not anticipated: one that the end user could arbitrarily pick and try. For example, if Windows XP was stated explicitly by the developer to run in the VM, could you try to run Windows Vista, or would the VM just refuse to even let you try? No guarantee of it working, but if the VM wouldn't keep you from trying, I would say you could run an arbitrary OS in that VM. All of this is opposed by the types of VM that require that the guest OS be ported or specifically changed to run in the VM.--MARQUIS111 18:36, 19 October 2007 (UTC)

It seems someone deleted this column, but I'd have to do a binary search to find when... I have a clear understanding of what it means. If an editor didn't understand and marked a yes where a no was appropriate, the correct response it to correct the error, not remove the column. For some platforms, one can expect that a new OS for a hardware platform it supports would be installable on the platform without special work. VMware, for example, allows this; e.g. Windows 7 installs and runs on VMware, even a version from before the first version of Windows 7 was created. Sure, it works better if the VMware tools are installed, but they're not required (and the ones for Vista work fine). Where custom programming is required you don't have a platform that supports an arbitrary OS. I think my definition matches MARQUIS', so I think the column should be restored. Alternately, "Runs Arbitrary OS" should be added to the "Gues OSes column. I propose the following definition for a Key/Legend :

• Runs Arbitrary OS: Can run a Guest OS without modifying it, and hence is generally capable of running any OS that could run on a physical machine the VM simulates.

--Elvey (talk) 21:25, 5 July 2009 (UTC)

Vmware entries grossly inaccurate

I noticed this missing column/info ("Runs Arbitrary OS") when I was wondering why the Host OS column for vmware is grossly inaccurate. http://pubs.vmware.com/guestnotes/wwhelp/wwhimpl/js/html/wwhelp.htm lists many OSes that are missing: MacOS (though Darwin is listed), OS/2, SCO. BeOS is a good example of an OS that VMware runs because it runs arbitrary OSes. Ref: http://74.125.155.132/search?q=cache:C-RNsvLywl8J:communities.vmware.com/message/719740%3Bjsessionid%3DF76FBD2F50EF4222D62249B4C1A7A11C+vmware+beos&cd=5&hl=en&ct=clnk&gl=us&client=firefox-a . Ditto: Windows 7 (beta and RC, at least). I feel safe predicting I could get Trustix, SEBSD, OpenBSD, etc to run under VMware as well.--Elvey (talk) 21:25, 5 July 2009 (UTC)

Oracle VM

Need to include the newly released Oracle VM. —Preceding unsigned comment added by Andreas Toth (talk • contribs) 00:47, 14 November 2007 (UTC)

Add new feature to the table?

What about seamless integration? See http://liquidat.wordpress.com/2007/09/05/screenshots-virtualboxs-seamless-integration/. I think that`s very good and worth to be listed as feature. —Preceding unsigned comment added by 84.183.148.33 (talk) 13:44, 6 December 2007 (UTC)

Here here, I second that. I know that virtual box and parallels support it as well as vmware fusion. This is a very compelling feature, and one I'd like to be able see highlighted in the features table. --Rgb9000 (talk) 23:03, 9 May 2008 (UTC)

Compare maximum number of LCPUs per VM and total RAM supported. —Preceding unsigned comment added by Pisapatis (talk • contribs) 13:07, 23 May 2008 (UTC)

Comparison of Method of operation?

You think it would be worth to make some site just with the theme comparisons of virtualisation methods?

Forum for virtual machines in general?

Great work you done here! You know a forum or newsgroups for virtual machines in general? Like with sub forums for all the most important emulators?

DOSBox

Perhaps it should be noted that DOSBox runs slowly deliberatly. The current way it's written makes it sound like it's a bad thing it runs so slow. Jawsper (talk) 19:51, 10 December 2007 (UTC)

DosEMU USB

It is possible to use USB mass storage devices such as Pendrives or HD's, by mapping it to a floppy drive (B:). —Preceding unsigned comment added by 87.205.229.16 (talk) 15:15, 8 January 2008 (UTC)

Hyper-V

Now that the veridian hypervisor has been released as hyper-v and the RC0 release included with server 2008, is it time to mention it here? —Preceding unsigned comment added by 216.124.224.116 (talk) 21:42, 22 April 2008 (UTC)

I agree that Hyper-V (Windows 2008) has to be added to the band-wagon. —Preceding unsigned comment added by Pisapatis (talk • contribs) 13:09, 23 May 2008 (UTC)

Needs Administrative Rights

We should write that the application "needs"/"does not need" administrative rights. --Ilhanli (talk) 20:48, 4 July 2008 (UTC)

File compatibility

I'd like to see some mention of this, with e.g. VMware has the ability to convert Parallels and Microsoft VMs but not afaik the other way around. It's mentioned in the article, but not yet here. 87.165.198.178 (talk) 16:22, 26 September 2008 (UTC)

And VirtualBox seems to support both without any need for conversion! How about that for pro-competitive practice? ;) 87.165.198.178 (talk) 16:24, 26 September 2008 (UTC)

colinux host OS

I personally have been running colinux on a windows vista SP1 host for several months with no difficulties, though according to the official colinux website vista is still unsupported *shrugs* --Kuzetsa (talk) 16:45, 1 October 2008 (UTC)

Hypervisors VS hardware emultators (sorting/reordering tables)

There's a lot of content in this page and I feel like additional sorting would make the page a better read. So I suggest to change the layout in any manner available to make the difference between hypervisors & "virtual machines relying on an OS" more obvious. For example they could be grouped(sorted) together.It seems to me that these roles are completely different, even if some software solutions allow both. What do you guys think of this ? Am I mistaken on something ?

BTW, this is just a suggestion and maybe it's not possible to do this for some reason, that's why I'm here asking questions. 82.121.214.9 (talk) 12:52, 13 December 2008 (UTC)

image type compatibility

Hell, I have no idea how to create a table. But I have some knowledge I want to share. There are many different image formats, proprietary ones and documented ones. It's always questionable which emulator is compatible with which kind of images. Rather it's questionable which formats can be already converted into another format which what kind of tool.

I would like to see a table for these kind of comparison so we can add more and more informations later.

VMware:

• vmdk (features: growing, shrinking, pre-allocated, defrag)
• physical disk

VirtualBox:

• vdi
• vmdk
• vhd

Bochs:

• vmdk
• raw / flat / .img
• physical disk

Qemu:

• Qcow
• Qcow2 (features: growable, compressed, encrypted, snapshots)
• raw / flat / .img
• vmdk
• physical disk
• http://bellard.org/qemu/qemu-doc.html#SEC15

VirtualPC:

• vhd —Preceding unsigned comment added by 87.172.179.56 (talk) 01:55, 3 February 2009 (UTC)

Something is badly wrong

Look at after the table, there is a load of garbage. At first look, I did not undesrtand where that comes from Porttikivi (talk) —Preceding undated comment added 07:43, 8 December 2009 (UTC).

Virtual Box administrative rights

It seems to me that needing Administrative Rights to install a package should not be counted as VirtualBox needing administrative rights.

Actually running VirtualBox and setting up new VMs is possible without being root/Administrator...

--Keeper of the Keys (talk) 17:05, 11 January 2010 (UTC)

Trango vs. VMware Mobile Virtualization Platform

I believe that the listing for Trango Virtual Processors should be edited to be VMware Mobile Virtualization Platform since the buy-out of Trango by VMware (2008). I believe most information is still correct, although support for host CPU architectures other than ARM have been dropped. See the MVP homepage, the old Trango site redirects here now. In fact, there is no wikipedia article on MVP at all, and it is not listed under VMware Software. 193.128.61.237 (talk) 15:50, 2 February 2010 (UTC)

Bochs is a virtual machine?

Bochs is a virtual machine? I was thinking that he was a emulator.187.89.197.232 (talk) 21:26, 5 March 2010 (UTC)

can boot guest from hard disk

where can i get information if a virtual machine can boot a physical OS as guest-OS?? Or is it "Can boot an OS on another disk partition as guest"? Then it should be renamed. --Txt.file (talk) 14:54, 16 May 2010 (UTC)

Plex86

Shouldn't Plex86 be in this comparison? Algotr (talk) 10:03, 19 September 2010 (UTC)

It seem like Plex86 never got ready for use. Algotr (talk) 10:17, 19 September 2010 (UTC)

Avanti/FreeAXP

I've added Avanti and FreeAXP to the General Information table twice in the past 24 hours, only to have them apparently removed. These are virtual Alpha emulators. I also moved VHDsoft to its correct location alphabetically. I am going to try a third time.

Bclaremont (talk) 20:13, 19 October 2010 (UTC)

I've cleaned out the list to contain only notable entries - that is, entries with existing Wikipedia articles. - MrOllie (talk) 20:24, 19 October 2010 (UTC)

So if I create FreeAXP and Avanti entries that describe the products, but do not advertise them, would they be acceptable? Would posting the product SPD's minus any marketing jive be okay?

Bclaremont (talk) 22:08, 19 October 2010 (UTC)

In theory, but if you do not have independently written sources that meet our requirements the articles could be deleted. - MrOllie (talk) 23:07, 19 October 2010 (UTC)

Updating

Lots of information here is going out of date fast. We really need a rewrite.Jasper Deng (talk) 06:26, 28 January 2011 (UTC)

Hyper-V 2008 R2 SP1 3D acceleration dispute

Previous discussion from User Talk

The following comes from User_talk:Jasper_Deng#Revert_on_Comparison_of_Platform_Virtual_Machines.

You reverted my edit on the comparison page, stating I had no source. I think you misunderstand RemoteFX's operating method, which exposes a virtual graphics processor supporting DirectX 9+ (9c?) to a virtual machine, in addition to improvements in the RDP protocol. I correctly cited that the accelerated graphics feature was only for guests that run *certain* editions of Windows operating systems, which is backed up by my citation. If you believe me to be in error still, please let me know and why, as I do not want to get into a revert war and understand the negative impact it has on Wikipedia. (Posting anonymously.)

RemoteFX is not a direct part of Hyper-V. Hyper-V itself doesn't support 3D acceleration. RemoteFX concerns remote connections, and by your argument all virtual machine software supports 3D acceleration, which is obviously incorrect. RemoteFX is technology to do 3d acceleration on remote connections; the virtual graphics adapter does not itself support 3D acceleration.Jasper Deng (talk) 02:05, 20 February 2011 (UTC)

You are plainly incorrect, as found here: [1]. Where are you getting your sources? Do you have experience with, or have you deployed Windows Server 2008 R2 SP1? (Which, by the way, is not yet available via retail channels, but is available to volume license customers.) Where is *your* source that I am wrong? If you cannot cite sources, and you will not revert my sourced edit, I will take you to dispute resolution.

Addendum—I will be restoring my change to the article some time tomorrow evening if you cannot back your argument up, probably around 2200 UTC.

You don't seem to be getting this-Wikipedia lists 3D acceleration as per local support, and RemoteFX is just that-remote, and that doesn't count under Wikipedia's article. Once again, please sign your comments!Jasper Deng (talk) 19:05, 20 February 2011 (UTC)

Wikipedia is a free encyclopedia that supports anonymous contribution, and I prefer to be anonymous. Moving on, you do not understand RemoteFX, obviously. There are two components to RemoteFX, and even a cursory reading of the Technet Library, Microsoft's blog posts on this issue (the Remote Desktop Services Team Blog, the Virtualization Team Blog) and other sources would provide you with this information. Simply put, there are two components to RemoteFX. First, there is a software stack improvement with Remote Desktop Services allowing hardware accelerated graphics to be sent over the wire, with optional hardware accelerated compression. Second, there is a hypervisor stack improvement that allows virtual machines to be given a virtual 3d graphics adapter, to allow non-physical operating system environments to access hardware accelerated 3d graphics. These are *separate* features under the same name, collectively termed RemoteFX. The former and the latter are separate technologies implemented separately in the OS. Were you to install Windows Server 2008 R2 with SP1, you would be able to test this out yourself and find that upon enabling the Hyper-V role and the Remote Desktop Services Role and the RemoteFX role service, you would find that under the Hyper-V management console you could add a RemoteFX virtual graphics adapter, and configure how much graphics memory, roughly, should be allocated to the virtual machine. It is not my fault that you do not understand these things and will not cite a source to back up your claims, and I will be reverting your unwarranted and unsubstantiated change now. If you revert my addition to the article again I will take you through dispute resolution.

First, concerning signing, please read WP:Sign your comments. Second, please do not make threats. Third, the 3D acceleration occurs remotely, not locally. It's local graphics that Wikipedia is doing there.Jasper Deng (talk) 02:18, 21 February 2011 (UTC)

Concerning your edit summary, I'd like you to know that you should not revert while discussing as per Wikipedia guidelines.Jasper Deng (talk) 02:19, 21 February 2011 (UTC)

Don't edit my comments in the discussion to you. And please stop citing policy incorrectly. Re: your claim, the graphics acceleration is happening in the guest operating system, via a paravirtualized graphics adapter. If you dispute that, please cite an article to that effect.

How am I citing policy incorrectly? The graphics acceleration is only happening remotely, and remote connections have always been a way to get 3D acceleration in VMs; thus, local graphics are the only ones that count. Concerning policies, you don't seem to be getting WP:Civility and WP:Assume good faith, WP:3RR and most importantly WP:Consensus, which you aren't achieving here.Jasper Deng (talk) 03:32, 21 February 2011 (UTC)

I am taking this discussion to Talk:Comparison_of_platform_virtual_machines and undoing your edit to my comment. I am additionally requesting dispute resolution assistance from Wikipedia_talk:WikiProject_Software and Wikipedia_talk:WikiProject_Computing. 03:49, 21 February 2011 (UTC)

New discussion

I am disputing the revert done by User:Jasper Deng on this article and would like discussion and consensus on my changes. I specifically claim:

1. My edit was reverted unnecessarily, as I cited a source and clarified my edit to the table with a reference.
2. The other editor, User:Jasper Deng has not cited one source to dispute my claims, while I have, in the meantime, added sources (as shown above).

I will be commenting under the name pseudonym for the purpose of this dispute. Pseudonym 03:49, 21 February 2011 (UTC)

You are misinterpreting your source.Jasper Deng (talk) 06:24, 21 February 2011 (UTC)

I disagree, as I linked to earlier, [2] plainly states, "The RemoteFX 3D video adapter enables 3D capabilities to a user connecting by using Remote Desktop Connection (RDC). A RemoteFX 3D video adapter can be added as a device to a virtual machine, which provides a Windows Display Driver Model (WDDM) driver with support for DirectX 9.0c." The virtual RemoteFX video adapter is attached to the virtual machine at the Hyper-V server, and is done so with the instructions that follow. That is 3D acceleration, no? That it is limited to certain guest operating systems (Windows 7 SP1 Enterprise, Ultimate, Windows Server 2008 R2 SP1) was noted. You insisted in the prior discussion that this graphics adapter was for remote use only, but that's a claim further than the cited article states, which is that RemoteFX enables 3D acceleration for remote desktop use. I believe it is against Wikipedia policy to cite yourself, otherwise I would perform the experiment on my home PC. Or would that satisfy you, if I took a screenshot of a Hyper-V VM on my PC with the virtual graphics card in the device manager list? You tell me. Pseudonym 07:12, 21 February 2011 (UTC)

Addendum; my original source additionally stated, "GPU Virtualization / GPU Virtualization is a technology that exposes a virtual graphics device to a virtual machine. RemoteFX exposes a WDDM driver with the virtual desktop, and it allows multiple virtual desktops to share a single GPU on a Hyper-V server." This clearly states that the virtual machines use a GPU on the Hyper-V server. Pseudonym 07:26, 21 February 2011 (UTC)

I think you're misinterpreting your source. That's all I think you're doing. See below comment (in Outside Comments).Jasper Deng (talk) 17:47, 21 February 2011 (UTC)

Outside comment

If I understand the situation correctly from a technical point of er view, both VirutalBox, VMWare and Hyper-V support 3D acceleration through the use of paravirtualization (i.e. guest-specific drivers). That, in the case of Hyver-V, these same drivers are also used for supporting 3D acceleration over a remote desktop connection would not change this situation. —Ruud 15:14, 21 February 2011 (UTC)

Jasper I think you are missing that 3D acceleration has to happen to happen on the actual graphics hardware (unless you're emulating 3D hardware, but then you can no longer speak of acceleration), so the graphics command have to be transferred from the guest to the host somehow. All virtual machines do this through paravirtualization and whether these paravirtualized drivers are supplied by the VMWare Tools or are already present in the OS to support RemoteFX matters very little. —Ruud 15:27, 21 February 2011 (UTC)

My issue is that it's not local and that remote connections have always been a workaround to 3D acceleration issues. Prove that Hyper-V provides local 3D acceleration.Jasper Deng (talk) 17:45, 21 February 2011 (UTC)

I think you completely fail to understand how 3D acceleration is implemented in virtual machines. You prove Hyper-V's implementation is different from VirtualBox' or VMWare's (it is not). —Ruud 18:09, 21 February 2011 (UTC)

I know how it works. Hyper-V however is unable to provide acceleration locally, thus it can't be counted. Remote connections, as I have stated, are often used to work around 3D acceleration issues in virtual machines. But now I propose a compromise:

We will not say style="background:#9EFF9E;color:black;vertical-align:middle;text-align:center;" class="table-yes"|Yes (the yes template), but instead, style="background:#FFB; color:black;vertical-align:middle;text-align:center; " class="table-partial"|Remote only.Jasper Deng (talk) 18:11, 21 February 2011 (UTC)

Your explanation is wrong so you clearly do not understand how it works. VMWare and VirtualBox do not do anything "locally" either. Their implementation works, technically, exactly the same as the implementation of the remote desktop connection. —Ruud 18:16, 21 February 2011 (UTC)

Jasper, the 3d acceleration occurs not on the remote client, but on the host operating system of the virtual machine. This is the same way Virtualbox, VMWare and other virtual machine clients do it. When you run Virtualbox and check "Enable 3D acceleration", what it does is add a virtual graphics card to the virtual machine. That graphics card typically has limited capabilities (low OpenGL or DirectX support) and is used for providing limited graphics support to the VM. That's what RemoteFX does. The graphics acceleration happens on the virtual machine host, and the output is compressed and sent over TCP/IP. RemoteFX is a server-side technology with support for decoding the updated protocol on the client. Nothing in the Remote Desktop Protocol 7.1 specification allows a client with 3D graphics to provide hardware accelerated graphics to software running on the host machine. Pseudonym 19:00, 21 February 2011 (UTC)

Here's an additional source: Remote Desktop Services Component Architecture Poster. This shows the server has the physical and virtual graphics processor, and all the client does is decode the data send over the remote desktop protocol. And I quote, "RemoteFX contains the following elements : Host-side V-GPU rendering and capture for 3D" Pseudonym 19:16, 21 February 2011 (UTC)

A screenshot would be the ultimate proof. The article is about virtualization in servers not remote desktop.Jasper Deng (talk) 19:19, 21 February 2011 (UTC)

Your comments make so little sense (see "not even wrong"), that I'm not going to continue this discussion with you. —Ruud 20:36, 21 February 2011 (UTC)

Explain this.Jasper Deng (talk) 20:38, 21 February 2011 (UTC)

I'm reverting back since this discussion isn't producing anything.Jasper Deng (talk) 20:44, 21 February 2011 (UTC)

You might want to have a look at WP:CONSENSUS first. The situation has been explained to you, in detail, by two other editors. You have only offered some nonsensical ramblings, indication you thorough fail to comprehend this subject. —Ruud 20:47, 21 February 2011 (UTC)

NOTE:This discussion has been moved to User talk:Ruud Koot.

No, it hasn't? —Ruud 21:18, 21 February 2011 (UTC)

Jasper, I think the misunderstanding here is because of two checkboxes under the experience tab of the "Remote Desktop Connection" program. The two options are "Desktop composition" and "Visual styles," among others I may be missing. Those settings, combined, allow the appearance of Aero to be used instead of the Windows Classic theme. This is not the same as 3D acceleration. Windows Virtual PC, the example you cited on Ruud's talk page does not support 3D acceleration. In fact, look at the table on the article, 3D acceleration for Virtual PC 2007 is a big fat template:no. RemoteFX, 3D acceleration != Aero. When you enable Aero for a virtual PC machine via RDP, what you're doing is having the remote desktop client render the Aero frames and do the composition of windows to do the Aero Glass(tm) effect. RemoteFX is entirely unrelated to that. Pseudonym 22:57, 21 February 2011 (UTC) —Preceding unsigned comment added by 173.18.86.157 (talk)

Prove however that RDP isn't the only way to get RemoteFX. RemoteFX is technically part of RDP, you know.Jasper Deng (talk) 23:00, 21 February 2011 (UTC)

It's part of Microsoft's Remote Desktop Services, but is not wholly included under the term "RDP", which is just the remote desktop protocol and is published here: Desktop Protocol: Basic Connectivity and Graphics Remoting Specification. If you download and install this, you'll note the only specification pertinent to RemoteFX is a codec used to transmit compressed bitmaps from the server to the client. 3D acceleration is not mentioned. The 3D acceleration is done on the host machine, RemoteFX the protocol enhancement and RemoteFX the virtual graphics acceleration technology are just confusingly named, that is, Microsoft uses the same name for two different things. The specification for the codec part Desktop Protocol: RemoteFX Codec Extension is also available from Microoft, and it states:

"The Remote Desktop Protocol: RemoteFX Codec Extension is an extension to the Remote Desktop Protocol: Basic Connectivity and Graphics Remoting (as specified in [MS-RDPBCGR]). The RemoteFX Codec Extension specifies a lossy image codec that can be used to encode screen images by using efficient and effective compression."

Jasper, if you have no sources, then please cite your technical qualifications for asserting I am wrong. I have cited five sources clarifying what, exactly, RemoteFX is. Pseudonym 23:41, 21 February 2011 (UTC) —Preceding unsigned comment added by 173.18.86.157 (talk)

You are misinterpreting them. Let me put it that way.Jasper Deng (talk) 23:47, 21 February 2011 (UTC)

VM Limits

I am constantly trying to update wikipedia, but my changes are reverted back.

1. I'm Technologov, long-term VirtualBox community member, and understand it well, along with some other virtualization software.

2. VirtualBox 4.0 does have limits. http://forums.virtualbox.org/viewtopic.php?f=24&t=41211

Maximum host physical processors (sockets) = 256 logical Maximum host cores per processor = same (256) Maximum host memory = 2 TB Maximum host disk volume size = no limit Maximum number of guest VM running = 1024 VMs Maximum number of logical CPU per VM guest = 32 Maximum amount of memory per VM guest, 32/64 bit = 16 GiB Maximum number of SCSI + IDE disks per VM guest = 4 IDE + 30 for others Maximum disk size per VM guest = 2 TB

3. Quick look at Hyper-V, also shows limits:

http://technet.microsoft.com/en-us/library/ee405267(WS.10).aspx

Please let me update stuff. —Preceding unsigned comment added by Technologov (talk • contribs) 02:26, 15 May 2011 (UTC)

Forums are not reliable sources. But, you should add Hyper-V's limits.Jasper Deng (talk) 02:29, 15 May 2011 (UTC)

Well, there are some hard coded limits in source code, here: http://www.virtualbox.org/browser/trunk/src/VBox/VMM/VMMR0/GVMMR0.cpp#L121

What is the procedure ? Should I speak with Oracle devs, and get the limit docs official ? VirtualBox _has_ limits. —Preceding unsigned comment added by Technologov (talk • contribs) 02:41, 15 May 2011 (UTC)

Use inline citations. See WP:Citing sources for help.Jasper Deng (talk) 03:05, 15 May 2011 (UTC)

Merger proposal

I propose that this page be merged with Comparison of VMware Fusion and Parallels Desktop. Both Fusion and Parallels are covered on this page, and I feel that their specific page serves only as an advertisement. In fact, many of the criteria on this page have been copied verbatim from the Parallels advertising website, including the criteria: "Keyboard shortcut (F6 or Fn + F6) for hiding/showing Parallels Desktop and all its windows." and "TimeMachine backups can be synced with SmartGuard snapshots, to reduce the space required for backups." The page is clearly biased and serves no purpose that Comparison of platform virtual machines cannot. As a result, this page should be merged with the redundant and biased smaller page to protect Wikipedia's integrity and decrease user confusion. Tutleman (talk) 20:07, 1 September 2011 (UTC)

• Oppose — the parallels vs fusion page goes into a lot more detail and deals only with Mac hypervisors. I would however propose that we:

• Move to Comparison of platform virtual machines for Mac OS X -- samj _in^out 06:56, 2 September 2011 (UTC)

• Oppose — I'm with samj on this one, rolling in this smaller article would lose a lot of Mac specific detail, it should be part of a larger Comparison of platform virtual machines for Mac OS X article.