Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization

At the Linux Foundation’s Collaboration Summit in San Francisco last week, executive director Jim Zemlin encouraged vendors and developers to standardize their virtualization activities around KVM—not Xen.  This whole Xen vs. KVM debate is getting annoying, but first off, this “news” isn’t news: there shouldn’t be any shockwaves from this late-in-coming statement.  Why?  Because KVM has been officially included in the Linux kernel since early 2007.  So it shouldn’t be any surprise that The Foundation (that has a nice Big Brother ring to it, no?) is encouraging KVM adoption. 

But the crux of the issue is really around why The Foundation chose KVM in the first place.  It’s simple: KVM = Linux; Xen = Xen (that is, a purpose-built hypervisor derived from Linux).  In fact, guess what KVM stands for?  Kernel virtual machine.  So if I were going to include one in the Linux kernel, which would I choose…hmm…  While that seems like an easy decision, there are consequences.  For starters, Xen is the more widely-adopted platform that already has abundant commercial support and an active open-source community.  Second, KVM is not a true hypervisor: it’s a hosted VMM.  I don’t know why the Linux zealots are fighting this one so much, but they argue it is a hypervisor because it gives guest operating systems direct access to the hardware.  Fine, I’ll make a concession: KVM gives Linux hypervisor-like functionality.  But, it’s not a true hypervisor because it requires Linux.  It uses the regular Linux scheduler and memory management.  So, if you want to load up KVM, you have to boot up Linux—there’s no two ways around it.

What really annoys me, though, is the The Foundation’s attitude towards virtualization.  Take this comment from a Linux coder at IBM, which didn’t come from The Foundation’s Summit, but pretty much sum’s up its general philosophy:

I trust Linux to run on my dvd player, my laptop, and to run on the servers that manage my 401k.  Is virtualization so much harder than every other problem in the industry that Linux is somehow incompatible of doing it well on its own?  Of course not.  Virtualization is actually quite simple compared to things like real-time.

Wow.  If Microsoft has ever been criticized for trying to make Windows do everything, then The Foundation should be criticized ten-fold since they're doing the same thing but under the auspices of being a "non-profit consortium."  Of course, the undoing of the Linux coder’s logic is the last line in his very own comment.  If virtualization is a “simple” workload, than why do I need to boot full-fledged Linux to get it?  More importantly, why would I want to?  While running virtualization as just another application workload on top of an OS might be fine for dev/test, it’s not fine for production workloads.  Servers that are part of virtualization resource pools are not going to be doing anything other than hosting guest VMs because I want to eek as much performance out of those servers as possible.  To ensure that, the only thing running between my OS workloads and bare-metal hardware should be a thin, purpose-built hypervisor.  Why would I want the performance hit and increased attack surface that comes with booting a whole operating system if I don’t need it? 

MS understands this, which is why Hyper-V is a true hypervisor as opposed to it’s Type-2 VMM predecessor, Virtual Server.  Dell and HP understand this, which is why they give customers the option of ordering servers with XenServer or ESX pre-installed.  Organizations understand this, which is why they’ve standardized around true hypervisors like XenServer, ESX and Hyper-V.  Everyone seems to understand this except The Foundation, which seems to believe that all workloads are best run as just another application stack.  They’re not.  Don’t get me wrong, The Foundation’s Swiss Army Knife mentality can be appropriate—and even ideal—for many types of workloads.  But if you give me a nail, I want a hammer—not some Frankenstein “utility” tool.


Posted Apr 17 2009, 12:48 AM by Brad Moczik

Follow Me on Twitter

Did you enjoy this article? If yes, then subscribe to our RSS 2.0 feed or

Comments

mudrii wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-17-2009 4:51 AM

It is interesting argument but with a flow > if I will run Linux as base I will definitely run Linux as VM as well same for Windows .

How many VM do you know hat runs other that a host OS ?

Brad Moczik wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-17-2009 6:23 AM

Hey mudrii, I'm not exactly sure what you are asking...

Kevin Bowling wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-17-2009 8:55 AM

I use Xen to run many all Linux guests.  It is a great platform.  What pisses me off is that Linux is ignoring Citrix' merge requests for the past 2-3 kernel releases, then blew about 4 of us off in the LKML when we asked why.

Yoashhh! wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-17-2009 9:49 AM

You don't have to run a full-fledged operating system... As much as you want a thin hypervisor you could satisfy with just the code needed to run KVM - and it's not the whole kernel.

What Xen and VMWare do is emulating hardware - which is quite an overload.

Because KVM allows direct access this cuts much of the overhead of having a hypervisor, even if it's not true and relies on a kernel :)

Moreever, KVM doesn't rely on a specific platform,

like the Hyper-V, Xen and VMWare.

You could get virtualization everywhere. This is not an issue for Windows but it goes perfectly well with Linux's portability.

And consider the administrative benefits of having a hypervisor managed as an operating system.

One last thing -- it's true that current commercial virtualization products are successful, but they all act the same basically. Here comes a second approach, which promises a whole lot of good things, and just because it wasn't tried before doesn't mean it's going to suck. The Linux Foundation has a chance to try an alternative, and I already mentioned a key factor to their choice - portability.

I think you try to measure them as a commercial company like Microsoft, and because the Linux Foundation is a different organization it has different goals - I won't be even surprised if commercial adoption wasn't one of them...

The Linux Foundation Abandons Xen After Microsoft/Windows ‘Hijacked’ It | Boycott Novell wrote The Linux Foundation Abandons Xen After Microsoft/Windows ‘Hijacked’ It | Boycott Novell
on 04-18-2009 4:22 AM

Pingback from  The Linux Foundation Abandons Xen After Microsoft/Windows ‘Hijacked’ It | Boycott Novell

The Linux Foundation Abandons Xen After Microsoft/Windows ???Hijacked??? It | All about MICROSOFT wrote The Linux Foundation Abandons Xen After Microsoft/Windows ???Hijacked??? It | All about MICROSOFT
on 04-18-2009 4:42 AM

Pingback from  The Linux Foundation Abandons Xen After Microsoft/Windows ???Hijacked??? It | All about MICROSOFT

nil wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-18-2009 4:56 AM

Ok explain something here Xen and V-hyper lack memory stacking, load blancing...  Basically they are bits of crap.  Lack the features to compete with VMWare.

Simply on the means to run multi-able clients well 2.6.31 Linux kernel with KVM will beat Xen and V-hyper completely.   Better memory management and better resource management.

ESX XEN and V-hyper.  All have kernels that run at the core of the hypervisor.  They all need min function to start up.

Linux kernel can be cut back quite a lot.  Final one is speed.    ESX XEN and V-hyper all require double switching to share virtual hardware.   IE switch to hypervisor then to host of device.

With host of the device and hypervisor as one you cut out switching.

VMM type-2 or VMM type-1 from secuirty is not much different.   Virtual Virtual was a poorly done VMM type-2.  It was still lacking all the features need to compete with VMWare.

Basically don't like it increase features.

There is a major advantage to KVM something you would love missing.  You need to migrate you don't have a KVM machine but you have a machine running QEMU.  Guess what you can migrate to QEMU on Windows that had not loaded a hypervisor.  Other solutions lack the flexibility to migrate where ever even if its not best.  Still running is better than not running.

Could Xen upgrade to KVM the standard virtual io interfaces KVM uses that it shares with lguest and qemu.  Yes.   In theory OS running inside could migrate between them all.

Xen is left behind because of many things.  They choose not to push patches mainline.  They choose not to work into integration with other programs.  blog.xen.org/.../hosted-xen-project-kxen-available  Yep xen is also returning to type II VMM.

anon wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-18-2009 7:26 PM

How dull.

If it isn't news, then why are you reporting about it like it matters?

Alan Fried wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 12:01 AM

This is kind of a weird article-- what are you complaining about? That the head of the Linux Foundation advocates Linux software? Stop the presses, that is scandalous. I do get tired of Windows fans acting victimized all the time.

Xen supports both paravirtualization and hypervisor. As paravirtualization requires specially-modified guests, and both Intel's and AMD's virtualization

support is commoditized, paravirtualization doesn't seem to have much usefulness. It's completely useless for Windows since MS won't cooperate.

KVM has a paravirtualization driver, and it also takes advantage of the hardware virtualization in Intel VT and AMD-V CPUs. So it runs unmodified guest and windows systems, and it runs them very fast. KVM is small and lean. Hmmm, sounds a lot like Hyper-V, so if Hyper-V is a hypervisor then so is KVM. Though KVM is missing the most important feature found in Hyper-V-- Byzantine Licensing Hell.

www.linux-kvm.org/.../FAQ

www.microsoft.com/.../hyperv-faq.aspx

Did you know that Hyper-V is based on Windows Server 2008? Pot, kettle?

Hint: Google before writing.

Wout wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 8:08 AM

I'm really tired of people speaking of things they know nothing about. The writer of this article has no clue as to what is important in these matters. Who cares were the stuff is handled. Big kernel small kernel type 2 3 4 5 6 or 100.

It's basic functionality, performance compatibility that matter. Who cares!

In my own tests I have found that KVM outperforms Xen and ESX 3.5. Management is OK, not great just OK. But I really really didn't like Xen in any way. Xensource/citrix's management interface is a lot better but technically the product blows.

I predict that Xen will die! Maybe KVM will  too. Hyper-v will rule the market on name alone. VMware will stay big. Xen and KVM will at the most have a nice market share!

Tsue Desu wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 8:57 AM

I do wonder how this author conceived this rant... He complains about the LOAD that a Linux kernel will put on a server, but forgets, to run MS Hyper-V, you need the entire Windows server installed and running. Heck, I'd take KVM over Hyper-V anyday... just because I actually do get the ability to run a bare metal virtual server.

Get real, stop Fudding about!!!

Andres wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 9:41 AM

No it doesn't have a 'Big brother' ring to it, 'The Foundation' sounds completely Asimovian, you know the organitation that was meant to replace 'The empire' with a better system?

So, if you insist on being derogatory, you'd better get a worse nickname.

Rambo Tribble wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 10:30 AM

Okay, so it seems you're saying Linux and Open Source are offering different avenues, choices as it were. Now, how again is that bad?

Brad Moczik wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 12:04 PM

OK, where to begin...  Apparently I touched a nerve despite the alleged confusion over what I'm writing about...  

People have posted several technical inaccuracies about each of the mentioned hypervisors, so I'll have to clear some of those up.  I suggest reading the articles I linked to so you can understand how they work.  For example, Hyper-V does NOT require "the entire Windows server installed and running."  If it did, Hyper-V wouldn't be a Type-1 VMM now, would it?  And that's just one of many misunderstandings that I'll attempt to clear up in a follow-up post.

So what exactly am I complaining about?  Well, like I said, that the Linux Foundation would recommend KVM is NOT big news.  Yet, anytime KVM gains any sort of ground or there's any sort of news, people inevitably bring up a Xen vs. KVM discussion in the sense of figuring out which will "win."  It's getting old.  

Second, I do take issue with the Linux Foundation's sole support for KVM.  While I understand their reasoning, I think the mentality is small-minded and not very compatible with enterprise-level thinking.  It is blanketly encouraging developers to focus on one particular hypervisor regardless of the product's maturity or whether it's the best solution for their customers--especially when the market is going in a different direction.  It's not like we're talking about Windows vs. Linux here: it's a Linux vs. a Linux derivative...

Bill Murray wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 2:11 PM

hi Brad,

         Interesting discussion, but I think you missed out my use-case.

For various reasons, probabaly bad,I run a Fedora desktop. I need a

EL4 VM, so I do it as a virtual machine. Xen need a special kernel which break

s AFS, but KVM is there, an works - it churns out 80% of the spped, in my application

of bare metal. So Yes, I am choosing one because it is in the kernel, but at that

perofrmance I don't need choice, I need one made easy...

          Bill

Damjan wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-20-2009 4:36 PM

Does Hyper-V have memory management, does it have a scheduler, does it have drivers for the hardware it's on? well guess what, instead of implementing all of those one more time (making the same mistakes/bugs or even worse different ones), KVM is using the Linux kernel for all of that, which means it gets a proven best of bread memory management, best of bread scheduler, drivers for almost everythings there is (just compare that with ESX).

BTW the linux kernel is 2MB on disk, the user-space part of KVM is 2MB too, you'd also need a shell and a libc... let's say in 50MB you can easily have a complete system.

Reading your rant, it seems to me you have no idea how KVM or virtualization in general works, past the marketing broshures you've read.

Adrian wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-22-2009 9:42 AM

First, Xen is not derived from Linux. Xen can use a Linux "parent partition" (or OpenBSD, or Solaris, etc). Xen itself has nothing to do with Linux.

Second, to claim that Xen has abundant commercial support is like saying that UNIX had abundant commercial support in the 80's and 90's. While that may be technically true on some level, but no two versions of UNIX were the same much like no two Xen implementations are the same today.

As for the Xen open source community, as you pointed out, participation in it is being discouraged.

You also keep going back and forth about Type 2 VMM. First you state that KVM gives Linux hypervisor-like functionality, then you dismiss it ..  because it requires Linux. I dare you to run Hyper-V or Xen without a parent partition and see how far you'll get.

You crap on KVM because it uses the regular Linux scheduler, memory management, power management .. oops, you left the last one out.

Further, your comments about attack surface are only valid if one can run Xen or Hyper-V without a parent partition. Otherwise you're just purposefully trying to mislead your readers.

Lastly, your comment about criticizing a nonprofit organization dedicated to fostering the growth of Linux is funny. Did you know, for example that the Foundation serves three purposes: to promote, protect, and standardize Linux. So blaming them for promoting Linux is like criticizing Microsoft for making a profit. Guy, wake up.

Toens wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 04-23-2009 12:39 AM

Either you make your "bare metal" hypervisor fatter and fatter by pushing more and more functions of an operating system in it - or you take an existing operating system and add a small hypervisor.

The latter seems a lot more elegant.

And the handling is a lot more elegant, too because you don't have to throw away all your existing solutions for clustering, loadbalancing, backup, monitoring and what not.

I can understand, that this is not compatible with the "enterprise-level thinking" of a software vendor like Microsoft, Citrix or VMWare (because it does not imply more sold software packages). But I guess, that it is compatible with customers who don't want to be dragged into another "ecosystem" of licenses, certifications and skill-sets they don't have and don't want to have (because it's not their core-business - for them it's just IT).

Regards,

Töns

Xen, KVM & the Linux Foundation | CHARGED's Digital Lifestyle at Work or Play wrote Xen, KVM & the Linux Foundation | CHARGED's Digital Lifestyle at Work or Play
on 05-13-2009 5:30 AM

Pingback from  Xen, KVM & the Linux Foundation | CHARGED's Digital Lifestyle at Work or Play

Xen vs. KVM: Ending the Debate | CHARGED's Digital Lifestyle at Work or Play wrote Xen vs. KVM: Ending the Debate | CHARGED's Digital Lifestyle at Work or Play
on 05-14-2009 6:34 PM

Pingback from  Xen vs. KVM: Ending the Debate | CHARGED's Digital Lifestyle at Work or Play

Brad Moczik wrote re: Xen vs. KVM: The Linux Foundation’s Small-Minded View of Virtualization
on 05-18-2009 11:06 PM

For those following the discussion, I recently posted an article addressing the feedback from these comments.  Check it out:

windowsconnected.com/.../xen-kvm-amp-the-linux-foundation.aspx

Add a Comment

(required)  
(optional)
(required)  
Remember Me?
Windows is a registered trademark of Microsoft Corporation.
Powered by Community Server (Non-Commercial Edition), by Telligent Systems Themed By nb development