Microsoft Hyper-V Leaves Linux Out In The Cold

Software, Cloud and Development Software & Tools
1

whitehartstag writes to mention that Microsoft has announced their new Hyper-V as feature-complete. Unfortunately the list of supported systems is disappointingly short.
“No offense to SUSE Enterprise Server crowd, but only providing SUSE support in Hyper-V is a huge mistake. By not supporting Red Hat, Fedora, CentOS, and BSD, Microsoft is telling us Hyper-V is a Microsoft only technology. More Mt. Redmond, Microsoft center of the universe thinking. That’s disappointing. Sure, if you are a Microsoft only shop, Hyper-V will be an option for virtualization. But so will VMware and XenServer. But if you run a mixed shop, Hyper-V won’t solve your problems alone — you’ll have to also add VMware or Xen to your virtualized data center portfolio. Or just go with VMware and Xen and forego Hyper-V.”

Link->Slashdot | Microsoft Hyper-V Leaves Linux Out In The Cold

2

Once again, FUD.

I have run Ubuntu on Hyper-V and it works just fine. All MS is saying is that SUSE is the only one they will support (i.e. you can call them up and ask “why isn’t SUSE working?”).

Don’t believe that? Well, look at this: Virtual PC Guy’s WebLog : Installing Fedora Core 8 on Hyper-V

One of the program managers on Hyper-V installs Fedora and blogs about it too…

And since when is SUSE Linux not Linux?

Note that officially VMWare (the main competition) supports SUSE, Redhat, TurboLinux, Mandriva and Ubuntu. You don’t hear Gentoo users screaming… because practically, they can still get stuff running.

I have heard that MS has already announced that RHEL is next in line for certification.

I guess that Solaris will probably be supported soon (MS and Sun have some interop stuff going on).

Disclosure: I have a vested interest in seeing Hyper-V succeed, as I will soon be working on it.

3

Isn’t KingKrool showing that he is a true MS employee ??

4

Nope. I am not an employee yet. And I gave you a cogent explanation. Why don’t you read it?

The problem with people nowadays is they refuse to actually experiment with anything and see if what is said about it is true.

I actually ran Hyper-V and checked that Ubuntu did indeed work. People have installed Fedora, Mandriva… blah blah blah.

Also note my disclosure. It describes honesty, as opposed to the article quoted above. The article is technically correct, but there is a lot of vague innuendo, and it loses sight of the fact that you can only certify a few OSes at a time (notice that Win98 is not certified). If you want to know something funny, I read that the script they have for Linux support in the beta version were actually for RHEL, and that is why the installation of the integration components in SUSE would go a bit roughly (certain warnings used to be thrown).

Also, the definition of support also depends on whether MS has integration components for a certain OS. Integration components are sort of like VMware tools in VMware server or workstation. And MS only certifies an OS if they are sure that the modules and all compile for the particular version of the kernel running in the guest OS… (you can run without the ICs, but it is not optimal and is not counted as being certified). Also note that unlike some virtualization products, you do not need to install the ICs (tools) to obtain functionality - so that means you can have higher resolution displays and networking without the tools installation.

5

^^^^+1 to ur post :slight_smile:

Working on a product that supports lots of different versions of UNIX itself (:P) i can tell u how much of a pain it is to certify on each n every different version, even if it ll run perfectly fine otherwise :wink:

But on the other side, NO commercial customer will use it in production if its not officially certified.

My 2 cents.

6

^I completely agree with your statement. No commercial customer will use it for mission critical non-supported systems. But a lot of what is being consolidated via virtualization is not mission critical. And many people will still do their own testing and find out that they have workarounds to get stuff working.

And my point was merely that the article states that Linux is being left out in the cold, when it is so clearly not. Most Linux distros work, and one is even certified. As time goes by, more will be. Note also that this is a server product - which means they will look to certify server versions of distros. So while OpenSUSE might never be listed as certified, don’t you think that it would work seeing as how SLES does?

As for BSD, let me tell you that MS is a lot more benevolent towards BSD… One problem with Linux is the GPL. MS doesn’t want to touch GPLed code with a ten foot pole. In such a scenario, they might actually depend on third parties to develop the integration components for them (the Citrix/Xen guys provide some support in terms of their hypercall adapter). They don’t really care about releasing the code for the integration components to the public (doesn’t really cost them anything), but what will stop some of the fanatics from claiming the entire hypervisor needs to be GPLed and claiming it is a derivative work (even when it clearly is not)? That slashdot article has an “informative” post which says that Hyper-V does not perform full hardware emulation - when it does!! In fact, it is probably the only VMM solution that does so. Try running VMWare workstation with an unsupported guest (or even a supported one) which does not have VMWare tools installed (Ubuntu has it by default now, but some distros might not). You’ll probably have problems with resolution (at least I used to, but I now only use Ubuntu as a guest and it works great). Whereas with Hyper-V, you don’t. You can run at high resolutions without installing the Integration Components… If people can’t see this, do you really think they’d understand anything about the architecture, and therefore the licensing?

7

Even if an OS fails to install/run under an MS virtualisation product, there are sometimes cool blog posts to work around the issues.

See Virtual PC guy

and Windows Virtualization Team Blog

For example this post on how to get Ubuntu working under VirtualPC avoiding some mouse/graphics issues: Virtual Varia : Making Ubuntu 7.04 Work Under Virtual PC 2007

8

GPL is a BOON to Linux, not a problem with it (from any perspective.) I am also looking forward to becoming a MSFTian soon, but that doesn’t imply everything else becomes a problem.

If VMWare wasn’t doing so well, MSFT wouldn’t have come out with Hyper V.

If Linux didn’t have such a strong hold in the server market, Windows Server 2008 would have been a much inferior product (competition is a big driving force for the market.)

If it is NOT derivative work, then it doesn’t hurt M$ to release the code with a non-GPL license (say Apache? MIT?) If it doesn’t contain derivative work, then how will them “fanatics” claim so? It was fun what M$ and “that other Linux wannabe company” were trying with the Linux kernel :slight_smile:

Warm regards,

Karan

9

How about looking at the statement I made in context? It is a problem for Microsoft if they even want to interoperate in a safe fashion. Why don’t you look at the recent GPL/non-GPL argument over ndiswrapper? It makes these issues complex, which means MS is obviously going to be cautious about how they proceed.

If VMWare wasn’t doing so well, MSFT wouldn’t have come out with Hyper V.

Interesting, since MS bought Connectix several years ago, and was actually looking to acquire VMWare (the price was too high in their opinion).

If it is NOT derivative work, then it doesn’t hurt M$ to release the code with a non-GPL license (say Apache? MIT?) If it doesn’t contain derivative work, then how will them “fanatics” claim so?

You seem to be confused as to what I said.

I claimed that hypothetically, MS has no real bone with the release of integration components under a GPL license. The things is, the ICs make calls into hyper-v. Is Hyper-V then a derivative work? Obviously not. But that doesn’t stop lawsuits from happening. Look at the Xen code and you’ll see that they’ve added a note to explain that calls into the hypervisor are NOT a derivative work! They had to go that far to clarify it. Releasing the ICs under any other license is no use - they would still be marked as tainted. As for fanatics, well the FSF tried to claim that MS had become a Linux distributor thanks to the GPL v3. I have no problems with the new provisions they added, but trying to claim that makes MS a Linux distributor is disingenuous (and I felt so at the time back when I never thought I’d be working for MS in any capacity).

Simple way to get around all of this - let someone else take care of the ICs. And they let XenSource do that. Frankly I’d guess that the ICs would ultimately be released under the GPL anyway. ICs are pretty brain dead code for the most part since the tougher work is done on the backend (i.e within the hypervisor and parent partition). MS has anyway released the specs for their hypervisor, so anyone can write ICs anyway. Or indeed, you can write your own version of the hypervisor from that spec (it is remarkably well specified).

If Linux didn’t have such a strong hold in the server market, Windows Server 2008 would have been a much inferior product (competition is a big driving force for the market.)

Where did competition talk come in from? Why are you going off track? BTW, FreeBSD existed well before Linux, and is arguably a better product with FreeBSD 7.0.

This article claimed MS is not supporting Linux on their product. I rebutted that argument. I explained why support for more Linux distros will take time. Why is this now turning into a more general debate on Linux vs Windows?