“As a major Linux vendor, one might expect that Red Hat’s new Red Hat Enterprise Virtualization for Servers solution would be able to run on Linux servers. You’d be wrong. Not only is that not the case, but the Management Server piece of RHEV, which provides virtualization management capabilities, requires users to be running Microsoft’s Windows Server. That’s no typo: A Linux vendor is requiring its users to run one of its key new products on the rival, closed source Windows operating system. According to Red Hat, the plan is to have a Linux version ready by some point in 2010. But in the meantime, Red Hat customers who want to run the virtualization manager must purchase or already own a Windows server.”
“Anytime we acquire a company, there is work to go from a proprietary code base to open source,”
1. VMware vSphere requires Windows.
2. RHEV requires Windows.
3. You can migrate from VMware vSphere to RHEV re-using all your existing infrastructure.
4. Profit!
Supposedly this was done because of time-to-market issues and they were going to think of a Java web based version and some Linux GUI tools later. Yes, the management layer was written for Windows, .Net, SQL Server and Windows Presentation Foundation. You didn’t read the requirements wrong. They’ve had a pretty reasonable amount of time to come up with something that runs on their own systems.
It really speaks volumes as to the state of producing any kind of user-facing GUI system on your average Linux system today and it was a very poor advert for Red Hat and Linux to people already using Xen or VMware.
I think that a large portion of the customers who want this kind of product would already have Windows systems in their environment. Especially if they’re already using an alternative like VMWare vSphere, which another post noted requires Windows too.
> They’ve had a pretty reasonable amount of time to come up with something that runs on their own systems.
They bought Qumranet 14 months before the GA release of the product. In that time they’d need to fix bugs and polish it to the point that’s it suitable for release – I’d expect a better first release from someone like Red Hat than a small startup. They’d presumably need to train support people and a whole bunch of other things too.
Not requiring Windows would either have meant rewriting the whole GUI bit (requiring a *lot* more development and QA), or suddenly supporting Mono and probably rewriting large chunks which use things Mono doesn’t implement.
Not being able to run the management tools on your own product is unforgiveable, and Microsoft must be pleased. To not even have come up with a preview is unbelievable. It really should have been a release blocker.
They’ve done a ton of work on KVM during that time but simply haven’t paid attention to the management tools in any way at all.
So they bought a enterprise virtualization system, which has Windows 2003 based servers and WPF (.Net) based user interface. It currently requires both the server and management console to be run on Windows machines.
And instead of benefiting from existing mono infrastructure, they’ll rewrite the whole system in another language (probably Java).
The only conclusion I can come up with is that RedHat really despises mono.
Certainly the fact that Mono is sponsored and controlled by their competitor doesn’t help, neither does the unclear legal status.
But all the politics aside, I fail to see how using Mono instead of Java would help them porting the whole thing faster. In either case they have to rewrite it from scratch, perhaps tweaking parts of the design along the way. And with all the Java know-how RedHat has, and maturity of Java itself, I’m not surprised seeing they chose it instead of Mono.
I think depending on an open source platform sponsored by a competitor like Novell is better than using a closed platform sponsor like MS.
Anyway, Mono has a very legal status; Microsoft published a “Community promise” (http://www.microsoft.com/interop/cp/default.mspx) when they say you are sure developing any implementation of several technologies, including C# and their base libraries; in other hand, Mono has developed a nice ecosystem with their own classes useful to develop new things from that with no danger of being sued.
Guys,
This is NOT a codebase that will work in Linux using Mono. It depends on too many MS specific technologies and was developed by Qumranet and used in production systems already with several customers before the Red Hat acquisition.
The effort to port it to using Mono is not going to be very different from the effort to rewrite it using Java and release it as open source with having to worry about “community promises” from Microsoft which doesn’t cover anything beyond base libraries.
My impression was a significant portion of the code was already written in C#, making porting almost trivial (Novell provides tools to automate Windows -> Linux .Net migration).
There might also be C# -> Java source code converters, too. But I don’t think it will be as efficient as using the same language / libraries across platforms.
Porting from C# to Java is equally trivial (after all C# was deliberately made very similar to Java). Just put two programmers that will rewrite the whole thing in a week or two. The difficult part is switching from one framework to another, as this requires more redesign, documentation and testing work, but that is also no better in Mono than in Java. I would rather say the opposite – Java being far more robust framework than Mono makes the difficult part easier.
In the end, why would RedHat _want_ to have one of their core products written in C#? I would not be surprised if they have ultimately moved from C# to Java even on Windows, to save some maintenance effort (that of course depends on the quality of the existing code).
Yahoo…… retrain the programmers from Qumranet because management likes Java/Linux better than Mono/Linux or .NET/Windows
A recipe for disaster for that part of the business. I have to wonder why they bought them in the first place?
If you buy a tech company just to wipe out all their tech then what is the point?
Qumranet brings original developers of KVM, Spice Protocol technology etc. A single management tool is just one part of it.
Maybe Redhat should look at one of their own sites:
http://virt-manager.et.redhat.com/
virt-manager does not scale to hundreds of thousands of systems. It is designer for personal desktops and workstations.