Please Sun, don't ruin VirtualBox
Unless your reptilian / reptoid / reptiloid overlords dispose otherwise that is.
As Sun Microsystems gets closer to the release of its xVM server virtualization platform, it's aiming to ease developers into the fold by spreading its lightweight desktop hypervisor around to hardware vendors. Sun earlier this week said that inked new multi-year OEM agreements for VirtualBox with three new businesses. …
Unless your reptilian / reptoid / reptiloid overlords dispose otherwise that is.
Virtualbox runs on Mac, yes. But it can't host OS X. Are Apple going to be including this Hypervisor?
Had to laugh - exactly how popular is just-about-zero market penetration? Xen and VMware are massively ahead both in product and the corporate mindset. Even Microsoft can talk a better pitch.
I gave VirtualBox a try - but gave up waiting for the thing to boot a virtual CD from a hard disc image, and attempting an OS installation. Solaris x86 booted from "CD" in just under five seconds or so with VMware, yet with VirtualBox it required minutes. Likewise, the install process was painfully slow. All this was on a 3.2GHz Pentium 4 with 2GB of memory.
No. VirtualBox is not there yet. It feels more like an emulator - a slow one at that, too - than a hypervisor. I'll be sticking with VMware for a while, I suspect. Don't get me wrong - I love Sun products (except for that behemothic monstrosity known as Java), but I also love a good product.
Sun would be better off *not* pushing VirtualBox right now, because to anyone who cares about VM (and therefore, has tried out the competition), will probably come to the same conclusions as I have: Nice toy, but hardly a patch on VMware.
They'll rewrite it in Java - runs anywhere, but soooo slooowly
Nice... Hands down, one of the best SNL shorts.
It's just not ready for prime time. In some ways I'm a little sad as Innotex has released many fine products in the past, including plenty of free OS/2 software.
Unfortunately Virtualbox is nowhere near as fast or as compatible as VirtualPC or VMWare. Windows supposedly works as a guest, as do some Linuxes. Unfortunately OpenBSD is extremely unstable and oddballs like Plan 9 manage to crash Virtualbox itself.
Using a combination of VMWare, VirtualPC and Qemu is currently wise.
VB lacks a paravirtualisation driver for disk access, making most disk intensive operations crawl, and that's even with the newly added SATA emulation in 1.6.4. Unless they get this fixed, I really can't see any serious outfits adopting VirtualBox over VMWare. Sure, GUI (display, mouse) response is great, giving the illusion of speed for some applications, but benchmark it and you'll be disappointed. I do wish them all the best though, as long as they don't do anything stupid like add a non-optional web-based management framework sucking up 100MB of RAM (hint, hint).
"Presently VirtualBox is available for free under a personal use license."
It is in fact available as an "open source edition" under the GPL, as well as under a proprietary licence with a few extras. For those comparing it to QEMU, the technology involved is actually based on QEMU, although they've taken certain concepts beyond what vanilla QEMU does, I believe.
I don't share Oliver's experience - no problems whatsoever in boot speeds, quick as you like. Been running a Linux based development VM quite happily on Solaris x86 and Mac OSX hosts and found it pretty fast actually.
The network implementation is a bit lacking however - no support for host-only networking - or at least without bridging interfaces which is a PITA on Solaris. Of course it's not going to replace ESX for production, but for playing around in dev it's a handy tool to have around and it is a much less obtrusive install than VMWare.
I want to run Solaris 9 on my hardware. Not Solaris 10 and not 11 either. I like 9. I can't exploit it and I can fully audit it unlike newer versions. Its sysadmin's can trace all its functionality. It doesn't use some non Unix "new way" to maintain it. I'll keep buying your kit as long as I can buy servers that run Solaris 9 and are no deeper than 21 inches to fit half decade old racks. Its easy to keep my business if you want it.
"VirtualBox is a hypervisor" it not really, its an emulator plain and simple, nothing wrong with that OC, but to be a true Hypervisor you NEED a realtime Hardware Abstranction Layer (HAL) at the core,each optimised to the hilt for the given CPU and its internal co-processing abilities.
for instance the PPC G4 linux codebase doesnt in any way take advantage of any massive Altivec hardware options, infact base PPC linux doesnt even look to see if theres an Alivec SIMD unit to use, its not like there doesnt exist a free and fast
libray that could be used by default and an agreement reached in the freevec C code
as for the X86 chips and their SIMD , while their MMX and SSE/2/3 are not as affective clock for clock as the Altivec or ther freevec C code, at least some core code has been somewhat optimised by a very few people, but that to NEEDS some serious work to bring it upto spec and get far more speed for a little coding skill and effort.
the problem is most x86 coders are of the well its good enough so i dont need to add these far better SSE/SIMD optimisations for an acroos the board speed improvement, and thats their failing, can t be bothered to start and finish a good idea and optimise the hell out of the simple core librarys and their Functions in cross plaform C as the frevec is doing for the PPC chips, a routine for the E600/Altivec, another tweeked one for the E300 that doent have ALivec but non the less can be optimised for that CPU to see a real improvements across the board, the same could happen for any good X86 codebase if it exists, any pointers on that btw?
I cannot at all second the comments w.r.t. speed problems. I run Ubuntu and XP as guests on a number of Solaris hosts, and they run fabulously. Out of the box. Only in Ubuntu I needed to add the user to a newly created group, and log off and on. :(
It does everything with close to native speed (that is well above 50%), except of DirectX.
And to Mac-guests: I bet Steve Jobs will sue the hell out of anyone who offers OSX-guests.
It's great that you've found a host/guest combo which works well, but...
If I was wanting a Linux guest on a Unix host, I'd use UML (user-mode Linux) instead (under Linux), and, based on experience I suspect it would be loads faster than VirtualBox or VMWare, even if it required a bit more tinkering to setup. The area where the performance really suffers seems to be Windows guest on Linux host, and (due to the low cost of the host) that's probably quite a likely scenario. I get no problems with boot speed, but certain 'real-life' disk-intensive operations really do suffer, as investigated here: http://forums.virtualbox.org/viewtopic.php?t=8119.
Fingers crossed they'll address this, because there are many reasons why I prefer VirtualBox to VMWare. We'll have to wait and see.
To the last poster, what on earth are you wittering on about? What a bunch of hard to follow, seemingly loosely related comments; are you showing off or trying (badly) to make a point?
Sentences begin with a CAPITAL letter by the way, but you're obviously more obsessed with assembly language than communication with non-virtual, non-silicon based entities.
even the spelling police like to use faster apps and VirtuallBox applications that are Optimised for better performance on a given hardware platform, that was the point incase your not clever enough to get it first time around.
if you look at that libfreevec library page linked, it seems clear enough, use that existing codebase or something like it thats made for your CPU of choice for free, and be happy people take the time to do these things for even the spelling polce like you to use.
not everyone speaks/writes english as their first language here, so whats the problem, as long as your clever enough to get the points they are trying to make for your benefit and greater education.
or you could just skip it and move along if its to advanced a subject for you to take onboard and use.
Yup, what you said :)
virtualbox is emulation it is not a hypervisor. Why does Sun do so many things half ass? They could be a good company if they just did a few things great.
Matt Bryant suggests, "Xen and VMware are massively ahead both in product and the corporate mindset."
SUN has been contributing significantly to the Xen project.
"The Xen Advisory Board... current members are Citrix, HP, IBM, Intel, Novell, Red Hat, and Sun Microsystems."
SUN released Xen (branded as "xVM Server") along side Virtual Box.
"Sun xVM VirtualBox is a hypervisor that Sun recently added to it xVM portfolio through the acquisition of Innotek."
"Sun xVM Server is an enterprise-class hypervisor appliance that serves as the basic building block within the Sun xVM platform for enabling datacenter virtualization and management."
Are you suggesting that one Open Source project contributed to by SUN is leading another Open Source project contributed to by SUN in "product and the corporate mindset"?
Virtual Box is targeted by SUN more as desktop virtualization while Xen/xVM Server is targeted more by SUN as server virtualization - they complement each other.
Xen and Virtual Box are really two different tools meant for two different purposes... comparing them the way you did does not make much sense.
Matt Bryant laughs, "exactly how popular is just-about-zero market penetration?"
VirtualBox seems like it is pretty popular among normal people.
"Sun xVM VirtualBox Breaks Five Million Download Mark... Currently downloaded more than 10,000 times a day, the new version includes more than 2,000 enhancements and full support for Mac OS X, Solaris and OpenSolaris host operating systems."
VirtualBox also seems pretty popular among trade journals.
2007.12 - http://www.techworld.com/opsys/reviews/index.cfm?reviewid=582
2008.05 - http://mac.elated.com/2008/05/30/virtualbox-16-review-free-vm-software-for-the-mac/
2008.08 - http://www.pcworld.com/downloads/file/fid,71402-order,4-page,1-c,utilities/reviews.html
2008.07 - http://www.itreviews.co.uk/software/s629.htm
2008.07 - http://blogs.computerworld.com/moving_to_virtualbox
2008.08 - http://www.macworld.com/article/134584/2008/07/mwvodcast59.html
2008.08 - http://www.eweek.com/c/a/Virtualization/VirtualBox-Delivers-Free-Multiplatform-Desktop-Virtualization/
While not perfect, it seems like a reasonable product, in comparison to the other competing products.
Seeing 10K downloads a day in conjunction with reasonable reviews from multiple trade journals placing it near-par with other competing vendor products does not seem to support your opinion.
Perhaps you could provide a reference for your "market penetration" opinion?
vbox is not an emulator, it's an hypervisor.. a type 2 hypervisor, like vmware workstation, get your facts right.
vbox is NOT based on QEMU, the only code it shares with QEMU is the dynamic recompilator, which is hardly used.
vbox has good guest OS compatibility... just try to run Plan9 or openBSD on VirtualPC...
sun will not rewrite it in java, that's a stupid comment... sun has not rewritten anything in java so far...
when you refer to "vmware" specify if you talk about ESX or workstation... if you talk about the first one, it's comparable to xVM server, not vbox... vbox is comparable to vmware workstation and does a good job at it... unlike other commenters, I have had good experiences with vbox
to clear out the last paragraph, vbox is not targeted at enterprise, xVM server is... so don't try to compare it to ESX
vbox is the 2nd most popular solution for linux users to use windows apps (first one is vmware player)... currently, market penetration does not only refer to licenses sold, but actual users... so the comment about "zero market penetration" is just baseless FUD, vbox is used by many... and xVM server is based on Xen.
Pam Courson states, "virtualbox is emulation it is not a hypervisor."
eWEEK Labs Technical Director Cameron Sturdevant disagrees.
"VirtualBox is a Type 2 hypervisor, which means that it is software that installs on top of the operating system installed on the physical host system. Type 2 hypervisors usually have slightly lower performance than Type 1 hypervisors that interface directly with the physical hardware, such as Microsoft's Hyper-V or VMware's ESX Server."
the real facts are right there in front of your eyes,its a PR Rebadged Emulator to make it sound cool, Emulator soundign old and tired, so lets all start calling it Type2 instead and make a load of new money from thoughs that dont know their cpmputer history....
now before you go on one, There is NOTHING WRONG with Emulation, infact much of todays current Emulations innovations you see today, came directly from the likes of the UAE codebase etc(Unix Amiga Emulator to use its Original historic name) and its innovative use of JIT (Just In Time)code compilation on the consumer x86 hardware.
Victor, you didnt make it very clear so ill help you out here, what you ment to point out i assume is, only VMware's _ESX_ Server is a true Hypervisor....as is the Sun xVM server.
all others are Emulators ;) or Type2 as the PR/sales teams wany you to call them
Type1 hypervisor = the real true meaning of the word "hypervisor" OC, being as pointed out above, running at the lowest possible level, AKA as a HAL (hardware Abstraction Layer), and preferably in realtime, that any OS kernel can then sit on top off and your OS sits on top of that.
Type2 hypervisor = the PR rebadged name for old school Emulators that sits on top of an OS , I.E 2 full layers ABOVE a real type1 hypervisor.
Type1 realtime HAL based being most peoples prefered option given the choice to get faster speeds on a given HW, but thats not currently very user friendly to your average user as yet, so Emulation/type2 is pushed as the better casual users option , and for good reason usually, as i said it works fine,.
and Vbox does have the destinction of being the only one that allows you to use US2 as a dedicated device to the Emulation ,good or making full use of US2 wireless cards and USB2 DVB-* TV cards.
no other Type2 lets you dedicate or even use USB2 divices AFAIK as yet...so good for Vbox evolution of the old code bases and ideas that came before it, now we need the ability to dedicate a real Ethernet PCi(e) to the Emulator too.
type2/Emulators QEMU, and the VMware range currently dont let you use USB2 (or at its full speed at least, i dont call USB1 usable)
Type 1: VMware's ESX Server, ....
Type2: VMware Server (formerly known as GSX), VMware Workstation, VMware Fusion,.....
I agree that the "type 2" designation is just a marketing stunt, but it IS different from emulation.
QEMU is an emulator, because it actually emulates the CPUs for the guest systems. On the other hand, vbox and vmware workstation virtualize the real CPU for the guests... meaning that they don't actually do emulation, they "just" redirect the instructions from the guest to the host in the assigned scheduling time for the hypervisor.
As a rule of thumb to recognize when you are dealing with a emulator and when it's an hypervisor, if the programs allows you to run machines designed for different CPU architectures than the Host's, then it's an emulator (like QEMU, that works on Hosts and with Guest designed for x86, x64, PPC, Sparc and others... other emulators include gaming consoles emulators, like ZSNES for the super nintendo, ePSXe for the PSX, and MAME for many different systems) else, if it only works with Guests under the same architecture as the Host's, it's virtualization (of any type), like VirtualBox, VMware workstation, parallels workstation, Xen and it's derivatives, LDOMs and others...
the "type 1" and "type 2" differ on whether the hypervisor is an OS by itself (baremetal) or not. For example, "type 1" x86 hypervisors include VMware ESX, Sun xVM Server, Xen, MS Hyper-V... type 1 VMM sit right over the hardware, meaning that they are the booting kernel (in Xen's case, it's the Nemesis-derived microkernel)... over this booting kernel sits a main domain (in Xen's language, a Dom0) which becomes the interface where you control the Guests or secondary domains (DomU in Xen's speech). Sun's LDOMs are the equivalent hypervisor but for SPARC systems.
OTOH, "type 2" hypervisors require an existing OS to work over it... examples include the ones mentioned before: xVM VirtualBox, VMware workstation and Parallels WS.
If there are any questions, please do ask.
It works perfectly for me. I'm running windows xp (because I am forced to for work) in VB on top of OSX. No problems. No slowness. It lightweight, reliable, and free. What more could you want?
An Anonymous Coward stated, "its a Rebadged Emulator to make it sound cool"
There is nothing I would like more than to see VirtualBox or xVM Server under Solaris able to be a real emulator, to be able to run SPARC Solaris under Intel... or MS Windows, Intel RedHat, or Intel Solaris under Solaris SPARC.
There are a couple of odd formerly compiled programs under SPARC that I would love to run under an Intel platform easily... not to mention some Windows WMI client calls to run on a SPARC in a container.
I have toyed with Wine and QEMU - hopefully they will be more robust in the next few more years.
That being said, I hope to see VirtualBox become a real Emulator instead of a Type 2 Hypervisor.
I don't know what all your comments about freevec are on about... If you look at the Linux kernel, you will find for x86 it basically uses MMX or SSE only for memcpy and I think calculating checksums for RAID stuff. It enables an Altivec bit (so apps can use it), and has Altivec emulation if an app decides to use it without checking if it's on the right CPU type. It's up to applications writers, GCC, and glibc beyond that.
None of this has much to do with virtualbox -- if they use altivec instructions in it, great. If not, the kernel can't help that.
you seem to have answered your own question ,or would have if you had read the Lebfrevec pages.
libfreevec appears to be an evolving SIMD routine replacement for GlibC , in this case optimised with Altivec code for any PPC CPU that has this SIMD ability.
just as you could replace the generic slower GlibC routines for any CPU that has other SIMD Co-pro abilitys, MMX/SSE etc with optimised versions there.
even GPU Co-processors if you wanted, and were willing to add to the work already done, such as the Gallium3D driver framework http://www.bitblit.org/gsoc/g3dvl/proposal.shtml with some imagination.
the point being made seems to be there are many totally or little used Co-Processors on many generic Chips and pluged into PC motherboards that could be used to give a massive boost to everyday apps by changing the generic glibc and related librays in whatever CPU family you are using today.
markos of that libFreeVec for instance tells you have to simply start using his replacements and get that needed boost
"...The next application you will load you will use the AltiVec functions in libfreevec! Enjoy! :)..."
"libfreevec 1.0.4 has been released a while ago, but it's only today that I've managed to finish with the benchmarks. Check the URL to see how a G4, a G5 and a MPC8610 compare to a Athlon X2 5000! :D
sure, its far harder to wrap 128bit Alitvec code into MMX/SSE SIMD, that the other way around SSE wrapper into Altivec native for Emulation etc but it could still be far better as you pint out above, they hardly even tryed to use the current hardware capabilitys to its SIMD potential
wgetis broken and should DIE, dev tells Microsoft