QEMU vs. VirtualBox

Why is it that the U.S. Government always releases a slew of RFPs just before Thanksgiving?  I’ve been swamped working on proposals since the third week of November, but we got the last one submitted just before Christmas so it’s back to normal (or what passes for normal around here) for a while.

I thought I’d take this relatively quiet period to do a quickie comparison between a couple of virtualization tools: QEMU and Oracle's VirtualBox.  For the comparison I chose to install virtual guest instances of Ubuntu 10.10 desktop from a downloaded copy of the iso.  The host system is an AMD 64-bit machine that is also running Ubuntu 10.10 desktop. Here’s the kernel version of the host at the time of this writing:  2.6.35-24-generic #42-Ubuntu SMP Thu Dec 2 02:41:37 UTC 2010 x86_64 GNU/Linux. I used VirtualBox 4.0 for this comparison.

Both products provide a graphical interface for building a new virtual guest instance.  Here are what a few of the QEMU build sequence steps look like:





 This review almost did not get written, because it took friggen forever for the QEMU install to complete.  It was only out of a somewhat morbid fascination that I let it go to completion.  I started the install at about 8:15am in the morning, and by 1:00pm it was finally finished.  By comparison, the VirtualBox install took just 28 minutes start to finish, and that included download time for updates,  since I had selected the option to do that at install time.  I did not select the update option for the QEMU install because I (fortunately) forgot to select NAT networking for it prior to starting the installation.

After the QEMU install was done, the Ubuntu guest remained equivalently slow.  Boot time was about 6 minutes as compared to 23 seconds for the VirtualBox Ubuntu guest.  Once I got the logon prompt from the QEMU guest it took a very long time to achieve the final QEMU result:


That’s right, a black screen. Which lasted so long I thought that was the final product.  About 15 minutes later, though, I got a login prompt, which promised to take forever to execute, so I finally put it out of its misery.

Oracle does not need to be too worried about competition from the QEMU camp just yet...

------------------------------------------------  UPDATE: 1/15/2011 -- The performance problem has been fixed by completely removing the kvm  and qemu-kvm packages and then reinstalling.  Thanks to all the comments which helped me zero in on the problem.  Also, at the suggestion of one reader, I reset the BIOS option for enabling virtualization and then power cycled the host.  Not sure if that was necessary, but I did it anyhow.
Performance of the VMs are now as others have reported, with install times taking minutes rather than hours.






Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Sırt ağrısı ve tetik

acemiyetsiz's picture

Sırt ağrısı ve tetik nokta

Bazen ağrı yerini ağrı kaynağı değildir. Tetik noktalar kaslarında küçük, sinirli solea noktalar olduğu diğer bölgelerde ağrıya neden. Örneğin, karın kaslarında tetik nokta, bel ağrısına neden olabilir ve geri başka bir yerinde ağrı neden geri bir bölümünde puan tetikler.

Iki ortak masaj yaklaşımlar tetik nokta kurtulmak nöromüsküler tedavi ve tetik nokta tedavisi uygulanır. Her tedavi, ancak kendine özgü yöntemleri vardır temel çatlak kremi tekniği ya kadar 30 ila 60 saniye kadar uzun süre tetik noktalar üzerine doğrudan baskı uygulamak ya da tetik noktalar üzerinde kısa, derin "sıyırma" vuruş kullanmak için.


Mark26's picture

The definitive test it is at final. Currently I can not comment much on this topic, I will keep my Opinions to this final. Sex Cams and this is for the visitors who are passionate about.

There’s a war going on here.

Anonymous's picture

There’s a war going on here. I’d like to comment but I’m not using Qemu but now I don’t know what to think with all the mixed reactions.

casino online australia


tolga drift's picture

hi you make share nice


Helen's picture

I was very encouraged to find this site. I wanted to thank you for this special read. I definitely savored every little bit of it and I have you bookmarked to check out new stuff you post.

Qemu can help you run a

Le Hoang Long's picture

Qemu can help you run a virtual machine when you do not have root privilege
I found it while trying to run some windows app;ications on my school library
I do not need root privilege to install qemu but virtualbox is impossible


ililnerede's picture


KVM/Qemu vs VirtualBox

Prudhvi's picture

I never had the problems you were talking about while using KVM/Qemu, it is a great product and works just fine. On the other hand VirtualBox is a great product for Desktop Virtualization.

Curious Qemu/KVM via cli vs gui

SwiftNet's picture

I have Qemu/KVM running at several client sites. The performance is very good. I do not use libvirt or virt-manager, I've always used the cli to configure the guests. I'm wondering if there is a performance difference between the gui and cli. I understand that the KVM/Qemu performance in this article is atypical, looks like the kvm module was not loaded, imo.

Where's the article?

Anonymous's picture

I didn't see any useful information in this "article". I agree with a few other posters that this discredits the magazine... I believe QEMU is aimed at more emulation than virtualization - look at the variety of processors and targets QEMU being used for, whereas VB tries to make a virtual x86 system up and going...

Oh, Me

BK's picture

You embarrass(ed) yourself, the magazine and Open Source in general with this incredibly lame article and your pass-the-buck attitude in the comments. Very immature and all hubris.



Vito's picture

It is more than obvious that you don't have any idea what you're writing about Doug. Your results equals my results when I was dealing with KVM for the first time. Instead of writing silly articles, why don't you study something about it to know how to troubleshoot the issues? If you want to see serious comparison of QEMU-KVM and Virtual-Box, go here http://www.phoronix.com/scan.php?page=article&item=linux_kvm_virtualbox4 , those guys did quite a good job without embarrassing themselves, good luck

qemu-kvm is excellent!

pcoq's picture

I am using Fedora and occasionally run Windows XP in a qemu virtual machine. I have fiddled with virt-manager a number of times over the year,s but I have never managed to get anything to run with it.

However, running qemu on the command line works perfectly.

qemu-kvm -m 1024 -boot c -vga std -soundhw all -localtime -hda xp.img \
-cdrom /dev/sr0

I haven't tried running just qemu -m 1024... Not really sure what the difference is, but I thought that virtualization is supposed to be faster with qemu-kvm.

Windows is slightly jerky, but usable without any great hardship. I have sound and internet connectivity without any problems.

The only problem I have not yet solved is how to share files between the virtual machine and the fedora host. I haven't given samba enough of a try yet to know how to make that possibility work for me.

I really like qemu a lot! It does absolutely everything I need and does it very well.

My Experience with KVM & VirtualBox

MikeJJ's picture

I currently use VirtualBox (on Debian Squeeze amd64) because of the Direct3D support now. However I used KVM in the past (with virt-manager).
From personal experience the guest installed and ran a hell of a lot faster using KVM. Installation time was about a 1/3rd of the time it takes in VirtualBox (and VMWare Server).

QEMU vs VirtualBox

BobbyC's picture

I could have written your article for you word for word. My experience was identical.

Is it a good comparison? Try VMware.

Anonymous's picture

I thought QEMU was a x86 PC emulator first, with optional virtualization added later. It shows. I had tried QEMU on Windows to run Ubuntu and openSuse. In both cases it was a slow experience.

VirtualBox was much better and faster, but Ubuntu didn't play well with QEMU or Virtualbox. QEMU had window/display issues with both Ubuntu and openSuse. openSuse does better under virtualbox.

Neither QEMU nor Virtualbox even come close to VMWare Workstation. If VMware is not available I'll fall back to VirtualBox. But I wouldn't use QEMU to do much more than run DOS.


renklisn's picture

I had my ass handed to me with some the comments on this article. Time to move on in a more productive way.


Virtualbox, it has waaay more features

Anonymous's picture

I prefer Vbox. It has soo much features. I can print from my virtual box to my windows printer from the guest machine. http://www.penguindoctor.com/2011/01/net-nootrality-strikes-early-in-2011/

I'll ask you for a favor

Eduardo's picture

Hello there!!

Well, we have here a very controversial issue about Virtual world.
It seems like this question brought back some other considerations into the arena. Things as - what's the best or adequate distribution to be used? Why Linus "A" is better than Linux "B"? Why a newer Kernel play worst than an old one? Along others.

So, Doug, I would like to propose a deeper dig in the matter. In order we can figure out what really makes difference in performance, with different hardware, SO and "virtual boxes"

Other way (my thought), we'll not have a conclusive look at the question.


Probably a good idea

Doug.Roberts's picture

Hi Edwardo.

Now that I've gotten the kvm kernel modules working I've been digging into virt-manager issues. One problem that I've found is that I have not been able to make sound work in the guests when starting the VM with virt-manager. I can get sound to work in the guest using command line kvm and passing the sound parameter:

kvm -hda /home/roberts/libvirt/images/Ubuntu.img -soundhw ac97

but that's a bit clunky.

Also, since the qemu/kvm VMs use vnc to display the guest I have not been able to find an equivalent why to do 2D acceleration like VirtualBox provides.

I imagine that there are a whole list of features that the more mature VM products provide that virt-manager doesn't. For example, both VirtualBox and VMWare make it easy for a guest Windows VM to see a Linux host's file system by running an embedded Samba server. You would have to manually set up a Samba server on the host to support local host file system access for a Windows guest using virt-manager.

Similarly, network management is much more full-featured in both VirtualBox and VMWare than it is with virt-manager.


Again, you're confusing

Dim's picture

Again, you're confusing kvm-qemu with VBox.
While VBox is a completely desktop-oriented product, not completely OSS and specifically developed by a team, qemu is oriented at multiplatform emulation and kvm is aimed at server virtualization, where multimedia capabilities are not as important as good drivers, performance and stability.

Having said that, I'd like to direct your attention at a project called Spice (spice-space.org), which, combined with qemu and kvm provides remote and local multimedia access much superior to VNC and RDP.

Thanks, but

Doug.Roberts's picture

I'm not confusing the two products, but I am comparing functionality of running virtual guests that have been created with these tools. Kvm may well have been aimed at server virtualization, but the qemu/kvm/virt-manager combination crosses the line between producing VMs intended for server use and user VMs. If kvm was intended primarily for virtual server use, there would no need for sound support, would there?

I am, with some justification, comparing the VMs produced with the virt-manager vm builder with those other two VM builder products, VBox & VMWare.

Having said that, thanks for the reference to SPICE. :/


> If kvm was intended

Dim's picture

> If kvm was intended primarily for virtual server use, there would no need
>for sound support, would there?

Actually, it's just something that came from qemu. KVM was developed by Qumranet, a company specialising in VDI, which also came up with SPICE. And as a VDI solution, multimedia support there was and is far better than anything VBox can do. This is why they stay local.

On the other hand, if you explore virt-manager, you'll see a lot of enterprise-y features, like support for centralised storage, live migrations, extra security etc.

So, again, VBox, having been concentrating ONLY on local desktop virt. has more features, and since it's not a community project, but something developed by a well sponsored company, it has been concentrating on local desktop virt. solely. Unlike libvirt, qemu, kvm and the rest of the API and stack you compare to it.

My point is, if you're comparing feature by feature, you have to list ALL the features, not those you specifically like, or find useful for only yourself, because that's just plain not objective. Otherwise it's like comparing a small towncar to a semitrailer, saying it's easy to park, and forgetting to mention the hull capacity figures.

This lack of objectivity is what drives the angry comments. You have already been proven wrong once in this forum, no need to keep dragging that on. So, no negativeness intended, I'm just asking you to try and be more objective

You're right

Doug.Roberts's picture

I had my ass handed to me with some the comments on this article. Time to move on in a more productive way.




Eduardo's picture

Thanks for your response!

And, PLEASE my concerns goes strictly in the way of collaboration.

Since the recent VBox company acquisition (what should change the way they use VBox in the business), also VMware is not Free and despite the fact of one VM's look like to handle better on Servers side Vs. final desktop users. It's quite relevant to compare those things and figure out what each one a doing better, faster, stable etc.


Same on Win XP host

kamelie's picture


I am trying to install Ubuntu 10.10 on a Win xp host and not luck either with QEMU manager 7.0.

Host machine:
Processor intel i7 (4 cores)

My virtual machine:
- 2 cores
- 1 GB RAM
- 40GB disk

After 2 hours I can not reach the end of the installation.

Anyone with good idea how to speed up Win XP host for ubuntu?

It seems it goes diretly to live mode without going to install mode ...

Only purchase hardware from ZaReason, System76 or other Linux HW

lamapper's picture

So many people have problems specifically related to proprietary hardware/software vendors with respects to Linux.

A friend of mine even gave Linspire a shot given their relationship, at the time, with Microsoft (about drivers) and got burned by the motherboard BIOS.

If you purchase your hardware from a Linux hardware (HW) Vendor it will ALWAYS work with Linux and Windows if you want to go there.

However the opposite can not be said to be true, if you purchase your PC from a Windows vendor / big box store, you will often experience problems installing Linux. Granted almost all of those problems can be resolved with patience.

Remember that there are more device drivers for Linux than any other operating system ever designed and/or released. The problems happen because many proprietary vendors wait 6 months to a year to release device drivers on new hardware. Still other proprietary vendors pay favoritism to Microsoft operating systems versus any other operating system. Such is the case with Intel (in the past), AMIBIOS, Linspire, Phoenix (BIOS), Nvidia (in the past), Foxconn (BIOS mother board problems for some models), Cisco / Linksys (new firewall/routers do not accept DD-WRT, OpenWRT or tomato firmware), etc...

To avoid problems related to proprietary BS, purchase your hardware ONLY from a Linux vendor like ZaReason or System 76. They do Linux right and avoid hardware that does not play well with Linux. (And as I said earlier, you can always purchase Windows separately and install it...knowing it will install, however when Microsoft drops support, and you know eventually they will, that hardware is 100% guaranteed to run Linux, probably every distro.

The best of all worlds - 100% Linux Vendors.

Special note: Avoid any vendor that plays lip service to Linux, but favors Windows to avoid hassles...ie. Dell (they bury Linux on their website, enough said), Big Box stores (all of them)

I only purchase parts and pieces from big box stores and even than I research their compatibility with Linux first to be safe.

If it starts up in live mode,

Anonymous's picture

If it starts up in live mode, there is usually an install icon on the desktop or in the menu. Live mode is great to make sure that everything works first before you install it. Make sure you have given it a virtual hard disk to install on.

If you are using XP, grab a copy of InfraRecorder and copy the cd to an iso. Mount and use that. It makes the installation a LOT quicker.

I like VirtualBox better than VMWare or QEMU. Just my preference.


tolga's picture

hello thank you have a very nice sharing


Rackmount's picture

Interesting that the internet brings out the worst in people. Doug reports his experience and people jump him for publishing that. There are civil ways to tell people you don't agree. Attack the premise, not the person.


Doug.Roberts's picture

Thanks, Rackmount. Good name, btw. :)

Sometimes writing about negative results brings out the angry side of folks. And providing an anomymous(e) comment forum allows some folks to feel that they can be far nastier than they would ever have the courage to be in non-anonymous interactions. There is an obvious reason that Slashdot calls them Anonymous Cowards. However, putting up with this social artifact is just part of the game.

But seriously, thanks for your comment.


"Sometimes writing about

john12345's picture

"Sometimes writing about negative results brings out the angry side of folks."

I agree.

"And providing an anomymous(e) comment forum allows some folks to feel that they can be far nastier than they would ever have the courage to be in non-anonymous interactions."

I disagree.

"There is an obvious reason that Slashdot calls them 'Anonymous Cowards'."

Ironic don't you think.

With that said i do think a national I.D is a good idea.


Sorry, but it just doesn't

Georgy's picture

Sorry, but it just doesn't work this way.

This article (published in Linux Journal mind you, not in a personal blog,
if it was the latter most of those who posted bad comments would not even care) pretends to make a comparison between an "open-source" (but company-controlled
with all sorts of schemes and tricks therefore not 100% open source) product
(VirtualBox) and a free software solution (Qemu/kvm) and concludes the article
saying "Oracle has nothing to fear" as if Qemu/kvm is completely broken/unusable somehow. The author also puts a bio that doesn't suggest "joe clueless layman".

This comes off as downright insulting towards the people who spent so much
time and effort to write such an excellent piece of software as Qemu/kvm
and in this community there can be no tolerance for that, sorry.

If the article was an objective comparison, pointing out the real advantages
of VirtualBox over Qemu/kvm in the desktop market (embedded rdesktop server,
working usb 2.0 support in the closed source edition, better support for
virtualized graphics, its recompiler that allows it to work much faster in
machines with no virtualization extensions, clipboard integration, guest/host
file sharing etc etc) then noone would say anything. Extra points if the efforts
underway in the Qemu/kvm community to compete and provide some of these features
were mentioned.

If the article said "I tried to make Qemu/kvm work but I couldn't. I
asked my LUG, on mailing lists and IRC and couldn't get any help
with my problem either. So, my opinion is that this software is
hard to install/troubleshoot right now and it would be nice to
have more documentation / better processes in such and such areas."
Again this would have been a useful contribution.

But this article comes off as "I couldn't get this software package
working right away because of my setup and because I like and use X
competing package instead, I will just declare that it is completely
broken and not worth considering." It is very reasonable to expect
lots of angry reaction to such behavior, since it is so unfair
and untruthful. Fact is, for a large number of people, Qemu/kvm
works fine (and even better than VirtualBox in some cases).

I 'm sure that most/all commenters have nothing personal against
the specific author, but at least I REALLY hope that all this negative
reaction will cause people to do actual journalism work before
posting articles to a magazine instead of posting
"personal blog-quality stuff" in the future. If one does post
such things, this reaction is what you can expect.
People will call on you.

Give us fair and objective articles and do your due dilligence
for now on please :)

Virt-io driver

yozshura's picture

This review is incomplete. Did the author use virtio driver for IO ? The driver increase KVM performance dramatically.

Truyen nguoi lon

Re: Virt-io driver

Andreas Mohr's picture

virtio drivers or not is not the core of the problem here I think, since I'm running a non-virtio setup (various VMs with older distros), and performance is not as abysmal as having to wait for hours for installation to complete.

However if there's missing virtio plus some other qemu setup foobar, then it's more likely that things will aggravate quickly.


Doug.Roberts's picture

Please see the update at the bottom article. Thanks for all the suggestions which helped me identify the problem.


Low quality article

Georgy's picture

Very low quality article IMHO.
Author should remove the "scientist" part from his occupation
since obviously there was no science at all used in this article.
"Blogger/Virtualbox fanboy" would be a suitable replacement.

"Long google searches" cannot help if you have no clue
about Virtualization (IMHO again). What you should do
is ask on the several mailing lists or IRC channels
available for ubuntu or upstream kvm/qemu or even
better ask for help at your LUG to get your setup right.
Reading the relevant pages at Ubuntu Wiki might also
help you. E.g., could it be that your user doesn't
have permissions to access /dev/kvm so virtualization
extensions were disabled? Only you can know.

I would also like to call nonsense on those who claimed
the problem is due to "Ubuntu's implementation of qemu".
Sure ubuntu's implementation even in 10.10 has flaws
IMO (like they didn't compile-in the support for vde,
or that they don't provide qemu-spice) but all these
are just details, not core problems.

We had (and still have) the extra annoying clueless
Ubuntu fanboys claiming Ubuntu can do everything
and now we also have the (just as clueless) "Ubuntu
haters" claiming that "Ubuntu is windows" and stuff
like that.

Took me about 15 minutes

SciFiDude79's picture

I'm certainly not going to dispute your findings, different machines act differently when using virtualization technology.

However, I just made a simple Qemu virtual machine (my first time using Qemu, BTW) with 1 processor, (2.50 GHz) 1024 MB of RAM and a modest 20GB virtual HD. I don't have an Ubuntu 10.10 .iso file but I'm using the Kubuntu 10.10 .iso which, as we all know, is basically the same OS. It took just about a minute or two to boot the Live CD, which is typical on my system and around 15 minutes to install, which is also typical. (including downloading language packs and whatnot) After installation, the system took about a minute to boot. (remember, Kubuntu boots more slowly than Ubuntu) That's about typical for my host system. I've been using VirtualBox for several years and I've alway found it to be a little shower than my host system.

The host system is an Intel Celeron Dual-core @ 2.50 GHz, 4096 MB of 800MHz PC2-6400 DDR2, 1TB Samsung HD and a 1GB Nvidia GeForce 9500GT graphics card. My OS is Peppermint ICE, which is based on Ubuntu 10.04.

Anywho, that's what I got.

I find your article rather unprofessional

Anonymous's picture

If you aren't able to properly install QEMU in your system it doesn't mean it's broken, as a matter of fact QEMU has thousands of satisfied users which use it on daily basis. Are they all fools?

You say you have done an extensive research on the topic, but from your article it seems you just installed QEMU, created a virtual machine, installed an OS on it, boot it once then decided it was rubbish. It took you a whole morning to install a virtual machine that boots in 20 minutes and you don't even wonder if you did something wrong.

In my opinion, if you weren't able to make it work you either don't have the necessary skills or your system have some incompatibility with that version of QEMU. In this case you could have done a lot better filing a bug report to the Ubuntu package maintainers to help them figure out the causes rather than writing such an useless article.

Tone it down, guys

Doug.Roberts's picture

I've been in the business for a very long time. I was one of the RedHat kernel developers before their IPO. I develop Linux applications for a living. I extensively researched kvm and qemu before writing this little article.

I know what I am doing, and I am reporting that something is broken in either qemu or kvm on a 64-bit Ubuntu 10.10 AMD system.

This isn't rocket science -- the process for creating a VM using virt-manger is dead simple. The fact is that on my system, which supports hardware virtualization, the kvm kernel modules simply are not working correctly.

So, my request is that if you can't be civil, either give me your contact information so that we can carry on the discussion off-line, or stop posting comments to this article.

You should realize that rude, anonymous comments don't carry much weight.

Oh, and they're unprofessional.

take the comments to heart

Anonymous's picture

This is an open forum, provided by the magazine to solicit comments to its articles. I won't comment on your appeal to authority or your request for personal contact information.

You must at least admit, based on your own comments, that you are comparing a working VirtalBox to a broken QEMU+KVM. Therefore, what is the purpose of this article?

KVM is not broken for many others using the configuration you present. You can easily verify this by looking at publically available bug trackers and forums. The performance of KVM has been tested by several reputable sources against competing solutions such as VirtualBox. It does not match your results. You can easily find this information by searching.

What, then, is the value of your article?

Had to reply to this, even

Dim's picture

Had to reply to this, even though I am not the one who posted the first reply.

There might be a problem with the kmods in Ubuntu, out of interest, do try this on another distro, a fresh RHEL or Fedora would be best.

I deal a lot with qemu and kvm, and what you describe is pretty weird, KVM based VMs, especially those running Linux or Windows with VirtIO drivers, are very snappy in my experience.

If you want to get performance to the max, use VirtIO for both network and disk, and set up the disk as RAW, not QCOW.

One additional thing, you mentioned you're going to test KVM as such in a future post, that kind of implies that in this test you were not using KVM, only QEMU, which is definitely not a fair test against another hypervisor, since on its own, QEMU is not a hypervisor, it's an emulator.
Lots of people understand that, and this is what causes all the negative comments IMO


Michael Reed's picture

I've always found KVM difficult to install and get working under Ubuntu. Like others, it's probably how I discovered VirtualBox.

UK based freelance writer Michael Reed writes about technology, retro computing, geek culture and gender politics.

i couldn't resist...

JShuford's picture

Interesting review. Thank you.

Have you considered shelling-out for VMware? I kinda like it!

...I'm not just a "troll", but also a subscriber!

I did

Doug.Roberts's picture

I actually ran VMWare for years, but switched to VirtualBox a few years ago because VMWare did not support fuse encrypted file systems at the time (maybe it supports them now, I don't know), but VirtualBox did.


Unobjective review

Renich's picture

This is wrong.

You shouldn't publish results just like that. Qemu works wonders on Fedora. IMHO, Ubuntu has a very bad implementation.

I have installed hundreds of VMs on Fedora and CentOS and, for a Fedora 14 VM minimal install, I take less than 20 mins to complete.

Besides, I think you're comparing these wrongly. VirtualVox implements parts of QEMU for it's virtualization; among other things (KQEMU, for example): http://en.wikipedia.org/wiki/QEMU#VirtualBox

Besides, it would be more appropriate to compare libvirt using KVM/Qemu, which is a closer comparison.

And, please, try something that's not Ubuntu as a Host. Any other distro will do.

It's hard to be free... but I love to struggle. Love isn't asked for; it's just given. Respect isn't asked for; it's earned!
Renich Bon Ciric


Well allrighty then!

Doug.Roberts's picture

Let's see if I understand the thrust of your contribution correctly:

Fedora good, Ubuntu bad.

Thanks for clearing that up!


Yeah; I also said: "IMHO"...

Renich's picture

Yeah; I also said: "IMHO"... I think.

But yeah, IMHO, Ubuntu sucks... at least regarding KVM/Qemu implementation. They've always favored VMWare, no?

Anyway, not my intention to start an Ubuntu VS Fedora flame. Let's keep the IMHO in mind, shall we?

And, besides, I offered a suggestion to the author so he could test on other distros. Hey, there's a hype going on for ArchLinux... why not there?

It's hard to be free... but I love to struggle. Love isn't asked for; it's just given. Respect isn't asked for; it's earned!
Renich Bon Ciric



JohnG's picture

I don't dispute the results you came away with, it's just disappointing that your tests turned out this way.
I have a fairly powerful machine (8 core i7-930 with 12GB of RAM), and at any given time I am running 6-8 VMs of various flavors using QEMU+KVM on an Ubuntu 10.10 base. I find the response time of the VMs to be quite snappy, the boot times to all be sub-1 minute, and performance of all VMs to be similar to native hardware. My VM stable includes Windows 2008Ent64 (running as a 4 core, 4GB machine), Centos 5.5 64-bit machines, Debian micro installs (using 1 core and 256MB RAM) Windows, and a pair of Win2003 64-bit servers running as an AD master and slave. I even run Office 2010 on the Win2k8 server with good results.
To emulate your test, I just did a from-scratch installation of a Kubuntu Lucid machine. Using Virt-manager, I gave it 1 CPU, 512MB RAM, and an 8GB virtio disk. The installation media is an ISO (kubuntu-10.04-alternate-amd64.iso) I keep on an NFS server linked to the VM host with GB Ethernet. Doing a manual install (where I needed to select the keyboard type, ok the disk partitioning, etc, etc) complete installation time was 54 minutes, 58 seconds from the "Create a new VM" mouse click to the "Remove installation media and hit 'OK' to reboot". This was a complete Kubuntu installation, with the full KDE SC suite, OpenOffice, fresh security updates pulled from the Internet, etc.
Boot time, from "VM power on" to a login prompt at the KDM screen is 8 seconds. After I enter my username/password, I get a usable desktop within 20 seconds.
It's a good bet the software on my machine and the software on your machine are the same. Somehow, somewhere, we have vastly different configurations that makes my system completely usable, and a joy to use at that, and your system a slothful morass of frustration.
My bottom line is, unfortunately, a common quote found on Linux Forums and help lists across the internet: mine works, yours doesn't. I feel for you, but clearly you're doing something wrong. :) I do take issue with your "qemu is not ready for prime-time" assertion; in my $1B/year company we use it in daily production on revenue-generating systems that run 24x7x365, and we do it with great success and cost saving.
Keep up the experimentation, contact me via email if you'd like more info on why mine works.


Very interesting

Doug.Roberts's picture

Could you clarify something, John? You said that you used the virt-manager to do the install, which is what I did as well. But you indicated that you specified a 8GB virtio disk. I thought you could only do that with virsh -- virt-manager does not support virtio in its interface.