VBox Linux guest performance vs VMware

[trubabalooey]

I'm wondering if any other users are experiencing slow performance of Linux guests in VirtualBox. It does not seem to matter which distro I use: Debian Wheezy, Fedora 17, Xubuntu (Precise Pangolin). Kernel versions are: 3.3.4-5 (Fedora), 3.2.0-32-generic (Xubuntu), 3.2.0-3 (Wheezy). I am using the latest version of VBox (4.2.2.0) on a Windows 7 x64 host.

Right from a VM start up it's slow: for example, I have plenty time to actually read the Grub welcome message as opposed to just seeing it flash by before it moves on to kernel boot selection; the rest of the boot process is equally slow. Also, properly installing Guest Additions does not make any improvement to performance.

For comparison purposes, I tried installing the above distros in VMware Player (9.0). The difference is dramatic! Load times are almost half the time it takes for VBox. For example, the time it takes from boot to login:
Debian Wheezy --> 70 seconds under VBox -- 36 seconds under VMware
Fedora 17 --> 75 seconds under VBox -- 37 seconds under VMware

So, before I give up the ghost with VBox, my question is...Is VirtualBox that bad OR am I doing something incorrect. I haven't used VirtualBox for almost 2 years now but I don't recall having any significant performance issues.

[Perryg]

post one of the guests log file as an attachment.

[trubabalooey]

post one of the guests log file as an attachment.

As requested...

[Perryg]

Code: Select all

HWACCM: No VT-x or AMD-V CPU extension found. Reason VERR_VMX_NO_VMX
00:00:01.416208 HWACCM: VMX MSR_IA32_FEATURE_CONTROL=0

There in lies your problem. There was a time when VirtualBox supported software virtualization and did so rather well, but for the newer kernels => 3.* there have been no updates to PATM which is the section that deals with this. I have asked here and been told that this is not high on their list of things to do and that I should (as I guess you should as well) use VMware since it does support software virtualization on the newer kernels.

See this for reference: viewtopic.php?f=3&t=51559#p236114

[trubabalooey]

Word up, Perryg!

Thanks for the info...learned more in your short reply than I did in a couple of hours of searching the 'net.
BTW, I agree with your statements in the linked posting. It is truly disappointing VirtualBox decided to freeze development on software virtualization. I, for one, preferred VBox over VMware; apparently, not anymore!?!

I guess I know what I'm asking for Xmas!

[martyscholes]

I guess I know what I'm asking for Xmas!

A newer CPU, I hope. I remember these same conversations in the 1990s during the 16 -> 32 bit transition where people understandably wanted to hold on to their 16 bit investments and lamented the 32-bit-only software, feeling abandoned. As it has always been with computing hardware, over time the vast majority switched to newer hardware, forgetting that they even had a problem.

Looking back, I recall the Hercules -> VGA trauma, the MFM -> IDE -> SATA trauma, DOS -> Windows trauma, SunOS -> Solaris (BSD -> SYSV) trauma, SCSI -> SCSI2 -> UltraSCSI -> SuperMegaNeatoSCSI -> Fibre trauma and so on.

Wow, I'm old.

[michaln]

It is truly disappointing VirtualBox decided to freeze development on software virtualization.

Well... how much is better performing software virtualization worth to you? Economic realities being what they are, for the cost of a new machine you can buy very little developers' time.

The fact is that in the last couple of years, it's been getting increasingly difficult to buy a new CPU without hardware virtualization support. I think I bought my last CPU without hardware virtualization in 2006. Needless to say, I'm not using it anymore.

There are still uses for software virtualization, but instead of mainstream it's now very much a niche usage. With nested paging and (on Intels) unrestricted execution in hardware, software virtualization can't beat hardware virtualization performance anymore. Hardware virtualization is improving bit by bit every year, so the gap is only getting wider. We can't ignore that, which means we have to work on improving our hardware virtualization support. Anything else would be crazy.

There is simply no business case to spend more time on software virtualization anymore. But if you have one, let us know (that's not meant jokingly). Or, if you have some time to spend on this, get the source code and see how hard it would be to improve software virtualization, the source is all out there.

[Perryg]

michaln,

I think your position has been well documented and while I am dealing with the fact that software virtualization is a dying beast in your position. I also have to tell others why they are having issues. If you can provide a better way to explain why please tell me.

As for a use case mine is rather unique and has no way of providing a monetary return. Since I retired I have been volunteering my experience to non-profits who are forced to use equipment that don't meet the VMX requirement. These people are not beggars but do not have the necessary cash to upgrade, at least for the present. I also support low income groups that try to teach people how to be enhance their computer knowledge in hopes they can get a job or improve their position if they have one and need to learn various operating systems.

My only regret is I should have spent more time on the programming side so I could in fact figure out how to implement the needed changes and provide them to the community.

[michaln]

I think your position has been well documented and while I am dealing with the fact that software virtualization is a dying beast in your position.

That's not our position, that's a fact of life We don't have exact numbers, but it's clear that the vast majority of our users utilize hardware virtualization these days. On server hardware, software virtualization was out several years ago; there were perhaps one or two Intel Mac models without hardware virtualization (if at all), and for a while only the very cheapest Intel CPUs did not include it.

I also have to tell others why they are having issues. If you can provide a better way to explain why please tell me.

I think you're doing a good job

As for a use case mine is rather unique and has no way of providing a monetary return. Since I retired I have been volunteering my experience to non-profits who are forced to use equipment that don't meet the VMX requirement. These people are not beggars but do not have the necessary cash to upgrade, at least for the present. I also support low income groups that try to teach people how to be enhance their computer knowledge in hopes they can get a job or improve their position if they have one and need to learn various operating systems.

Well... you can't squeeze blood from a stone. You have to ask yourself whether it really makes sense to run the latest and greatest software on outdated hardware. That has little to do with virtualization in general - installing Windows 95 on a 386 did not result in a good experience, just like installing Windows 2000 on a Pentium didn't, just like installing Windows 7 on a Pentium III doesn't.

Again, the reality is such that for the cost of upgrading to virtualization capable hardware, you can get maybe 10 minutes of developer's time, and probably not even that (you can buy an AMD-V capable CPU for $40 nowadays). I'm not saying that's good or bad... that's just how it is.

[Perryg]

you can buy an AMD-V capable CPU for $40 nowadays

OK but you need to go out and buy one of those processors and actually get it to work.

In reality it will take a new mother board and probably newer ram and power supply because the CPU in and of itself will not work either.

If I might suggest an easier solution. Add reference or some explanation to the users guide stating that continued upgrade to software virtualization is suspended or negated. As it stands people that do read this are being told that it is the finest thing since sliced toast and when they can't get it to work in a reliable way they come here with problems.