Slashdot Mirror
VirtualBox 2.1 Supports 64-Bit VM In 32-Bit Host
Stephen Birch writes "Following closely behind the mid-November 2.06 release of VirtualBox, Sun Microsystems has released version 2.1. This has a number of new features, but one of the most interesting is the ability to run a 64-bit VM inside a 32-bit host. Another useful feature is integrated host-based networking; no more fiddling around with network bridges. Sun is really giving VMWare a run for their money."
.. opengl acceleration on windows guests on any opengl capable host! beat that vmware!
Would a 32 bit emulating a 64 be like a fat man in the 200m dash?
I can't justify purchasing a 64 box for the house, but a beefed up 32 running 64 virtually is just the ticket to get SAP on Linux up and running. Merry Christmas to me!
One ring to bind them - should probably have more fiber and less rings in their diet.
here's a howto install Virtualbox 2.1 in Ubuntu Linux:
http://ubuntuforums.org/showthread.php?t=1015045
Another interesting new feature is the experimental 3D acceleration via OpenGL. From the manual:
With this new feature, if an application inside your Windows guest uses 3D features through the OpenGL programming interfaces, these will not be emulated in software (which is slow), but instead VirtualBox will attempt to use your host's 3D hardware.
This works for all supported host platforms (Windows, Mac, Linux, Solaris), provided that your host operating system can make use of your accelerated 3D hardware in the first place.
The 3D acceleration currently has the following limitations:
1. It is only available in Windows XP and 32-bit Vista guests with the Windows Guest Additions installed.
2. Only OpenGL acceleration is presently available in those guests; Direct3D is not yet supported and will be added in a future release.
Aside from a clunky GUI, the thing that stood out the most for me about VirtualBox is the abysmal snapshot support. Both VMWare and Parallels allow for a snapshot tree where you can instantly jump to any powered-on machine state that you have saved. VirtualBox, on the other hand, seemed to only support a linear, multiple-level undo.
Anybody know if any progress has been made in this area?
I have found that VirtualBox is a perfect alternative to VMware's expensive Workstation product. Before a friend told me about VirtualBox I was using VMware's Server free product as to how Workstation was meant to be used and not as a server side virtualization solution as VMware expected. So as soon as I checked out VirtualBox I dumped the ever-so-getting bloated Server program suite. I did previously pirate Workstation a couple of years ago before the free Server got released and decided I would try to go legit at that time which made it easy since Server and Workstation were compatible with each other on virtual machine files. As for Workstation product its ~$200 price tag is just way too expensive for my taste.
Now I'm using VirtualBox and I really do like it a lot. It seems to even be less resource intensive than VMware's offerings. Now the question is has anybody tried, or even if possible, to convert a VMware virtual machine to a VirtualBox machine?
This space is not for rent.
I have an Athlon64 but run a 32 bit OS. I tried running a 64 bit virtual machine using VMWare Server 1.0.x a year or so ago and it worked. The performance was not noticeably poor.
So... assuming I haven't missed anything too obvious, my response would be "No, vmware is not getting a run for their money." Not today anyways.
was the reason why I tossed out Virtual Box.
It was prone to problems, and became so annoying I ended up buying a license of VMWare.
There is also one area which is very unstable - OpenBSD support. It crashes the latest versions of OpenBSD, reports out-of-disk errors etc. OpenBSD is definitely more picky on the hardware it runs due to its strong security features, which Virtual Box doesn't appear to implement properly to make it look "real enough"
Sun has recognised problems with OpenBSD but has said its so far down the important-list it won't bother for some time.
Does this mean if I have more than 4G of memory the client will actually be able to use more memory than the client can see? So I can have a full 12G client on a 16G host that only sees/uses 4G of it?
I've been trying out VirtualBox for a while. VMWare had recently updated to v2.0 and had some annoying problems with the new tomcat based web front-end. It was unusable and drove a lot of people to other options. This was why I'd looked at VirtualBox.
It is easy to install and runs most OSes as a host. I tested the last two versions on CentOS 5.2 on 64bit and 32bit. The 32bit version running on my Inspiron E1505 laptop had issues with CPU utilization. No matter what was running (or not running) in the guest, it would completely spike the machine to 99% utilization. Fiddling with the CPU virtualization settings and other BIOS features had no effect.
Anyhoo, VMWare released an update that fixes the Tomcat issues. Xen is running great. Right now I don't have a lot of reason to switch, but VirtualBox does look very promising.
It means that the virtual network adapter can get IP connected without resorting to NAT. This was usually done by bridging a physical interface to a tun device and setting that tun device as VirtualBox's network device. Setting up this bridge requires using a script outside of VirtualBox to get everything set up. Now VirtualBox can do it from the GUI with no scripting required. In short, one can dedicate a physical NIC to VirtualBox by bridging it or allow VBox direct access to the host NIC.
The easy way to do networking with virtuals is to use NAT to pass TCP traffic to the virtual from the host's IP connection. That suffices for web surfing and other apps that don't severely exercise networking but it doesn't work well for things like VPN clients.
Virtualbox doesn't just give VMware a run for it's money, it's considerably superior in many respects:
- It's open source
- The gtk interface is at least as good as vmware's gui
- It's considerably faster on my system (no hardware virtualisation), windows xp boots in about half the time in virtualbox than in vmware, and applications generally open/run much more snappily.
- It's considerably more stable (on linux) than vmware is. In my experience vmware crashed about 30% of the times I used it, I even got a total system crash once that needed a hard reset (I think due to problems with compiz?). It uses quite an intrusive kernel module that creates a lot of latency in the kernel. This manifests itself mostly as skipping audio when audio is playing. Virtualbox has none of these problems, it's rock solid stable and doesn't hog the cpu like vmware does.
- Virtualbox seems to need less ram than vmware, I only have 1GB of ram in my laptop and swapping was unbearable with firefox and vmware open, yet firefox, virtualbox AND other applications can coexist fine with only limited swapping.
That's all the advantages I can think of of the top of my head, the only disadvantage I can see is that vmware supports USB devices whereas the free version of virtualbox doesn't. Other than that, virtualbox is just better all round.
Sam
This is how the loudness war is killing music.
If you wanted your VM to have an IP and appear as if it is a real machine on the network many people used to have to follow the 100 odd lines of documentation here: https://help.ubuntu.com/community/VirtualBox#Networking
Now they can just start it and it works out of the box.
I have to disagree. VirtualBox has the makings of a good product, but right now is too buggy and unreliable to be used in a production environment.
Most of the unreliability that I've encountered stems from virtual disk management. For example, if you have a virtual machine with a CD ISO mounted, what happens if you stop that machine and delete the ISO? This:
VM cannot start because the hard disk '/home/seizurebattlerobot/.VirtualBox/VDI/Windows Vista.vdi' is not accessible (Could not access hard disk image '/home/seizurebattlerobot/.VirtualBox/VDI/Windows Vista.vdi' (VERR_FILE_NOT_FOUND)).
Result Code:
NS_ERROR_FAILURE (0x80004005)
Component:
Console
Interface:
IConsole {ddc6fda1-a435-45ca-b43d-f9e88746e53e}
The only way to get the machine into a usable state again is to manually edit the virtual machine definition, which is a lot more complex than one would immediately think. Just look at the VirtualBox bug tracker for some horror stories.
The disk snapshot feature is also a mess and can result in data loss if you are not extremely familiar with how the underlying implementation works. The GUI dialogs that control snapshots are poorly documented and are definitely not production quality.
It is also not possible to shrink a virtual disk that uses snapshots. Normal GUI based activity has resulted in an inconsistent snapshot tree state that has caused data loss for me numerous times.
In short, I would recommend VirtualBox for anyone that wants a virtualization sandbox to play around in. To anyone concerned about data integrity, hates troubleshooting obscure and difficult to track down error messages, or wants to use disk snapshots at all, I would recommend waiting a few years before considering VirtualBox.
That's the truth. Sun, Xen and even Microsoft are giving VMWare a run for their money nowadays.
There's one interesting thing which has struck me, that I haven't seen any comments on. Namely, that VMWare is stuck competing between Microsoft on the one hand, and several Open Source projects on the other (with some of the Open Source projects having serious financial backing).
Being positioned between Microsoft and Open Source generally hasn't been a good spot to be in (indeed, has anyone succeeded here?). So I have to wonder how VMWare is going to stand up in the future?
I've been a big fan of VMWare in the past, as it has saved my butt more than once. However, now I find myself using Xen more, and seriously considering Sun's offerings.
To VMWare's credit, they have arguably the best person in the world for the job as CEO (at least on paper). Some might remember Paul Maritz as being one of the top people from Microsoft, as well as having led Microsoft's original *NIX strategy (I.e. Xenix). So if there's anyone who can compete there, it is him.
But still, it is not an enviable position to be in, and it makes me wonder how they are going to compete in the long term? Especially since, from a technology basis, the Open Source efforts are arguably better.
Anyone care to add some insightful comments on this? The only way that I can see VMWare winning is if everyone else screws up. While that's possible, there's a lot of money at stake in the Virtualization field, and I think the odds of that happening are low.
The best way to predict the future is to create it. - Peter Drucker.
Thus far, my virtual boxes have all been on a private network. I'm not even sure if they see each other, though I've not really tested that. I'm not even really sure how to open up the guests to the public network, though I'm 100% positive that it can be done. It's just that the defaults are all pretty secure.
That all means that your host is acting as a NAT router (by default anyway) and thus all the firewall that the host has will protect the guest(s).
Yes, if your guest gets infected, it's inside the firewall. Though, like I said, I'm not sure it can see the other guests, just the host. However, it's fairly easy to solve: turn off the VM, and roll it back to a clean state. I mean, if you're paranoid enough to be worried about such issues, you'll have old states which are known-good to roll back to. However, I've turned off pretty much all of WindowsXP's protections because it's hiding inside my Linux box, behind a cable-router (another NAT). The ability for something to get in and infect it is pretty much nil. Especially as I don't use IE or Outlook inside there (I use kmail for email, and firefox and konqueror on Linux for browsing, so no need) either.
Sun has taken a beating lately, like everyone has, but when I look at its massively multicore Niagara and Victoria Falls systems I see real viable breakthrough in the area of massively parallel computing. With Solaris Containers on Sparc I can take an old production Solaris 8 ( or 9 ) server and literally drop it into a single core of a Niagara machine and then make the old box vanish into a puff of 1U smoke without losing anything. Heck, the new machines will run 256 threads at the same time with no time slice issues. With VirtualBox on x86 we can now park almost anything from the x86 world into a SunFire x4440 ( 16 core AMD Opteron slayer ) in 2U of rack space. The theory, that I would love to test in practice, is that you can make four racks of older gear vanish into 6U of rack space with the SunFire x4440 ( AMD Opteron based ) and the Sun T5440 Server ( 32 core and eight floating-point units per processor ). That would be 256 simultaneous threads all running in one server and 16 cores of AMD Opteron in the other. And that means Linux/Windows and Solaris all running in two machines. I may be wrong but Sun has a hell of a grip on the future multi-threaded world.
...the glaring problem that I have with VirtualBox, VM management.
I love that they use XML for defining a system. I love that they have a virtual disk manager. But they do not have a process for bundling both together if I have a virtual machine with a disk image that I am dedicating to it.
The end result is migrating a VM or even upgrading an install with non-default settings is a huge hassle. And the default puts VM-client xml files in a different location than their virtual disk images.
In VMWare, I just move the entire directory and start a vm with the appropriate vmx file.
have they resolved the pain in the ass issue with using USB on Linux Hosts?
The only way to get the machine into a usable state again is to manually edit the virtual machine definition, which is a lot more complex than one would immediately think. Just look at the VirtualBox bug tracker for some horror stories.
This confused the hell out of me the first time it happened on a virtual CD mount. But it only took a few minutes to realize that all that needed to be done was to disable the CD from the GUI. It should be just as easy to disable a hard drive.
While it is bad form to refuse to boot over something so trivial I don't see this as a show stopper.
Disclaimer: I'm not using VirtualBox in a production environment.
But the ability to take multiple, branching snapshots is worth the price of admission alone. Version 6.5, which they pushed out a short while ago, has a new featured called "Unity mode", which basically takes programs running in the Guest and draws them on the Host so they act like any other program running on your host.
If you are a developer who uses virtual machines every day, $200 is a bargain for a tool like Workstation.
SUN of a BITCH....
Previously: "Linux... Toward the Sunrise..." Now: "Linux... Toward the-- No, now, part of Every Sunrise"
As far as I can tell, it's just a free as in beer version of free as in speech + beer QEMU.
How long before it comes out that virtualbox actually is QEMU, with a Sun sticker on it?
CPUs and GPUs are designed to handle *very* different kinds of instructions. GPUs are meant to work with matrices that are streamed into them. CPUs are designed to take a single stream of instructions and run it efficiently. While CPU-type instructions can theoretically be run on a GPU, you aren't going to get the kind of performance that you're dreaming of. Sorry, it just doesn't work that way.
It is pitch black. You are likely to be eaten by a grue.
I had the same experience and resolved it with disabling the disk in the GUI too. As you mention, it would be nice if VirtualBox would at least try to boot the VM even if a disk or two went on vacation (especially if it's a CD/DVD).
I've found VirtualBox to be much more pleasant than VMWare Player. Mainly due to:
- not so hard to get it working on 0-day kernels
- really annoying VMWare keyboard bug
That keyboard bug might be fixed now but the way VMWare (didn't) handle it is enough to move me on to more pleasant pastures.
http://www.virtualbox.org/wiki/Virtualization
Check this out, it is not a emulator.
You can't run FreeBSD in VirtualBox because it forgets to clear some flags after an interrupt, which causes FreeBSD to notice that the hardware is in an invalid state. I don't believe this is as serious with other guests, but they may get some weird behaviour from drivers. It's been a known issue since 1.x and still isn't fixed. Someone wrote a patch for the FreeBSD kernel that clears these flags, but it's far from an ideal solution.
I am TheRaven on Soylent News
Just SSH into your guest and run the apps remotely. You can run an X server on OS X or Windows. For a local machine it's even adequate to use unencrypted UDP on the loopback interface. You're likely to end up with about the same overhead as running an emulated 3D device, but without needing any special configuration.
I am TheRaven on Soylent News
VirtualBox can have the same kind of new kernel issues as VMWare. I went over tracking one of those down my blog (and this month went over working around the limitations in VirtualBox that prevent you from cloning a snapshot image).
I've found the stuff that VirtualBox has issues with straighforward to work around for the usual reason that makes open-source software easier to cope with: when I do run into a bug or limitation, it's sometimes possible to poke at the source code to figure out what's going on. In that snapshot cloning case, a quick read of CloneToImage and its associated code gave me a decent idea what was going on. That's why I run it instead of VMWare player: given anything close to feature parity, I'll take a slightly buggy program I can see the source code to over one that's closed.
The title of this post is rather misleading. It implies VirtualBox can run a virtualized 64 bit machine on a 32 bit processor and VMWare cannot. Neither of these are true. It can now host a 64 bit guest VM when the host OS is 32 bits.
Support for 64 bit VM's under 32 bit host OS's has been standard in VMWare's entire line ever since they included 64 bit guest support. Even the service console through ESX 3.5 is a 32 bit VM (Though it's not really fair to call it the "Host" OS)
AFAIK, virtualizing 64 bit guests does still require Intel VT or AMD Pacifica support on the CPU regardless on all products that support 64 bit guests.
Another way to put it is that it *emulates* the behavior of those APIs.
You use that word a lot. I don't think it means what you think it means.
Are you talking about a saved state or just a straight boot?
Under the influence of Post-Cyberpunk Gonzo Journalism
VMware has been running 64bit VMs under 32bit host OS (on 64bit hardware) since.. 2006, even before?
I'd say look at which products people are willing to pay for, and you know who is giving who a run..
I joined two users too late.
It's a known issue. Try to build world, or portsnap update and you are likely to see it. The bug was filed against 1.6 and was still reproducable with 2.0.6. I can't produce it on 2.1, because now my FreeBSD VM aborts during the boot process (as in, the VM aborts, not the OS - the window disappears and the VirtualBox GUI reports the VM state as 'aborted'). This is with an OS X (Core 2 Duo) host, but the eflags bug has also been reproduced on Linux.
I am TheRaven on Soylent News