Does VirtualBox VM Have Much A Future Left?

[rikkinho]

hm

virtualbox works well for personal use, i use it for test OS and a windows 8 guest when i need some app or test something, 3d and 2d is not great but it works

[cynyr]

There is really no need. KVM is way forward, it is GNU/Linux's killer app. It works now on ARM, Power and x86, while VirtualBox is x86 only due to hideous amount of x86 assembly code. Windows port of Vbox is of no use for me, I don't use Windows for years.

Oddly for me, as time goes on i spend more and more time on windows. Professionally i have to use windows. Autocad 2014, Excel, and all of the windows only vendor product selection programs are windows only. Also there is no Autodesk design review for linux. Wine might work for some of it, but being able to load a DLL into excel and call functions via VBA is required for what I'm doing at work. So I'm using vbox at work to have a linux guest, and a windows dev VM. I'd have liked to use KVM+libvirt, and virt-manager but it doesn't work on linux. There is no way I'll get vmware paid for by work, so that's a non-starter there.

[darkbasic]

The only thing I know of that can do it is qemu with KVM.

Xen too.

[duby229]

And what real use case for that? It seems to be slow for hardcore gaming and not really needed for anything else. So it looking fancy ... but that's all.

And if someone in mood for "real" 3D in VM it seems to be better idea to use more than 1 GPU in their system, detach one from host and completely give it away to guest. This way it could be more or less simiar to "real" GPU (still subject to some overhead). Sure, it takes specific setup: system where you can detach at least one PCI-E GPU and it also should be more or less recent hardware (with IOMMU). But wish to have powerful GPU in VM is exotic as well

I wouldn't call it exotic. It makes sense for anyone wishing to play video games on windows through a VM. Right now the only way to do that realistically is to have a second video card, with a second monitor hooked up to that. The only thing I know of that can do it is qemu with KVM.

EDIT: Even then if you have an Intel CPU that had the Virtualization capabilities gimped, you may be SOL.

[darkbasic]

And what real use case for that?

Fast compositing for example?

[SystemCrasher]

VirtualBox is currently my only shot at even using Linux at work. If I had a Linux host, I wouldn't need virtualization at all.
And the thing it does better than VMware is: it's free.

Well, if someone needs Linux IMO it is better idea to go on and use Linux. Both on host and guest. And maybe it sounds harsh, but usually nobody haves goal to give you some free solution of your personal problems. There could be some exceptions but they're rare.

In open source world people usually chewing on tasks when they're either interested in result or if they're just having fun doing something particular. In proprietary world people are working just because they're being paid for and so goal is usually to squeeze some bucks from you. Sometimes it also happens some opensource devs are doing jobs which are needed by someone else (e.g. large company). In this case company can hire that dev and pay for continuing doing that job. Which usually sounds like relly nice option for dev.

And so Vbox. I do not see how it fits. These who needed Linux likly mostly gone Linux on both host and guest and interested in KVM development. Then there is Oracle. I do not see any reasons why they could be interested in VBox. Its not part of their business and it's not like if Oracle is known in doing things "just for fun". Basically I do not get who will develop VBox and why they will do it. Probably that's why it got stuck, just as news article describes.

[SystemCrasher]

Qemu doesn't support 3d acceleration (with the exception of vga passthrough), so it isn't a viable alternative. Virgil3d seems pretty dead.

And what real use case for that? It seems to be slow for hardcore gaming and not really needed for anything else. So it looking fancy ... but that's all.

And if someone in mood for "real" 3D in VM it seems to be better idea to use more than 1 GPU in their system, detach one from host and completely give it away to guest. This way it could be more or less simiar to "real" GPU (still subject to some overhead). Sure, it takes specific setup: system where you can detach at least one PCI-E GPU and it also should be more or less recent hardware (with IOMMU). But wish to have powerful GPU in VM is exotic as well

[aksdb]

I simply fail to understand why someone would use VBox. Can it do something better than competitors? For me it looks like if Oracle mostly abandoned project and there was no community to continue development either (unlike in case with MySQL or LibreOffice).

VirtualBox is currently my only shot at even using Linux at work. If I had a Linux host, I wouldn't need virtualization at all.
And the thing it does better than VMware is: it's free.

[darkbasic]

Qemu doesn't support 3d acceleration (with the exception of vga passthrough), so it isn't a viable alternative. Virgil3d seems pretty dead.

[SystemCrasher]

VirtualBox has a fairly complete API for programmatically creating environments. libvirt nicely replaces most of that though. It doesn't seem like it should need a ton of work, it's always been a little rough, IMO, it's heavier than VMWare.

Are people leaving it because it's not updating every 3 weeks? Or is it not updating every 3 weeks because fewer people use it?

As I told, for me KVM is clean winner due to being built-in feature of Linux. Also, if I need one-shot VM to try some advanced setups, and is is not a group of identical VMs, easiest and fastest thing is to launch VM instance from command line, giving all options I need by entering command line. QEMU got load of options, so its possible to configure VMs very precisely in direct way. Yet, I do not have to resort to hardcore programming for such one-shot tasks. But sure, libvirt is good to have on my side, too.

Then KVM got another obvious advantage: it is good to run Linux on Linux (to provide extra separation/area for system experiments which can break OS). Linux kernel of most distros inherently supports "better" KVM drivers ("virtio"). Which are working much faster than fully-emulated virtual hardware. In most other cases you'll need some "additions", "tools" and other custom crap. But when you run Linux on Linux, guest usually well aware how to use these better interfaces WITHOUT any extra tools installation. So guest runs at decent speed you expect from modern VM solution without messing with installing "additions".

Also, virtualization is on rise. If someone fails to find any new feature to add... this likely indicates their project got some serious issues. Since Oracle is well known for screwing up MySQL (forked as MariaDB) and OpenOffice (forked as LibreOffice), I have reasons to think VBox is about to face similar fate, too. Except nobody from Linux side would bother self to fork when there is already KVM in mainline Linux kernel. And KVM got a lot of traction - just google something like "cheap VDS", "cheap VPS", "cheap cloud VM" and get idea: yeah, this technology is not going to disappear anyhow soon and got serious market demand. Not something Vbox can afford.

I simply fail to understand why someone would use VBox. Can it do something better than competitors? For me it looks like if Oracle mostly abandoned project and there was no community to continue development either (unlike in case with MySQL or LibreOffice).