Friday, September 04, 2009

RHEV - The RedHat failure not soon to be forgotten

Lets take a moment to look at Open Source Virtualization, there's Xen, OpenVZ, UML, KVM, and VirtualBox. Now for the datacenter you're really only looking at Xen or KVM (though OpenVZ seems to have a little bit of a following in the datacenter, its not an "Enterprise" offering from the big companies). Xen is on its way out, lets be honest, it has moved further away from the upstream
kernel as time goes on and Citrix is Microsoft friendly so people will embrace the FUD. What's left? Oh, KVM.

KVM is a really cool concept, I'm a big fan of turning the linux kernel into a hypervisor. That's just a really cool idea and has apparently proven itself to be quite useful and in some reports it is claimed to be faster than Xen so all the better. RedHat buys Qumranet, good move, now what? Lets write an Open Source Virtualization Suite that rivals the likes of VMWare ESX/ESXi + vSphere such that everyone in the world can enjoy the benefits of virtualization without being bound to closed source software from companies like VMWare. Again, good move. So what next? Lets force our Open Source faithful as well as all our customers to run a Windows box in order to use this pleasant administrative interface that front ends our completely Open Source Virtualization Environment. Wait ... what? You're kidding right? Nope.

So the office where I work has a site license for RHEL, lets just say that we weren't a mixed OS environment which would be completely feasible seeing that we have a site license. So you're trying to tell me that my CIO cut a FAT check to your company for a site license of an Operating System (as well as a nice repository full of software) and now I have to go elsewhere for another Operating System (that is closed source no less) in order to run your next generation Virtualization Suite? Kiss my ass.

If I were to roll out Open Source Virtualization tomorrow on RHEL, it would be RHEL 5.4 KVM + Convirture[0] not RHEV because I don't have a Windows box in sight.

I've heard rumors floating around that the plan was "get to market fast, port to Java later for cross platform". That's awesome, but in the mean time you're pissing people off and last time I checked its not good to make those who pay your bills angry. I'm not giving up on RedHat, I still really respect them as a company in many many ways (many thanks for all the sponsoring of Fedora!) but I feel like they really dropped the ball on this one, come on guys:

"We will be the leader in Open Source Virtualization" -
Brian Stevens (CTO & VP, Engineering)

Yes, you will but you need to work on moving the management end to an Open Source platform so your statement doesn't seem loaded.

Thanks.

rant.end()

-Adam

[0](http://www.convirture.com/index.php)

17 comments:

Jef Spaleta said...

Do you have a citation you can provide that will back up the rumor?

-jef

Anonymous said...

What is your source?

Jef Spaleta said...

I'll be a little clearer as to why I want to know the source of the rumor.

My understanding is that the windows only codebase that you are referring already existed in some form as a Qumranet owned codebase before Red Hat acquired Qumranet. My understanding is that it's an inherited codebase and not a codebase that Red Hat developed from scratch in-house.

I think its a little bombastic to suggest that Red Hat stop development on an inherited codebase. Isn't it far better that Red Hat is working on making cross-platform than just dropping the codebase entirely? Red Hat has a tradition of freeing codebases that they have acquired.

You really have to look back and understand what Qumranet's business model was before Red Hat acquired them to get the full context of what Red Hat purchased...including pre-existing windows only codebases..and not just KVM.

Adam Miller said...

Jef: My source of the rumor about the switch to Java was just from talking to people who were at the RH Summit.

I do understand the concept of inherited code and that's all well and good, but I think the cross platform migration should have happened pre-release. Why release a product that won't even run on the Operating System you ship?

Frank: Source for which part? Essentially all my information is from info piped to me from the RH Summit.

Jef Spaleta said...

Here's a fun fact.... the spread of misinformation can be modeled like the spread of a contagious disease.

Do do you feel worst about spreading the flu or spreading inaccurate information?

Let's have some fun... let's find patient zero. Let's find the person who started this rumor spreading.

Jef Spaleta said...

Adam,

Is the windows only codebase "released." I thought it was in beta and not actually released as a product yet. Can you point me to an announcement concerning the codebase in question being offered officially as a released product?

-jef

mether said...

"
I do understand the concept of inherited code and that's all well and good, but I think the cross platform migration should have happened pre-release. Why release a product that won't even run on the Operating System you ship?"

Because, the profit from the release can be used to fund the rewrite of the codebase to be cross platform?

Because it is not shrewd business to wait for a complete rewrite to make your first release when the virtualization market is highly competitive?

It would be nice if the Qumranet product was not written for Windows but it was. No amount of wishful thinking is going to change that.

Adam Miller said...

Jef: http://www.redhat.com/rhel/virtualization/ <--- You are right, RHEV is not official yet but you're splitting hairs at this point. They announced it at the Summit as Windows only. If they magically port it to Linux by official release date then I will gladly admit I was wrong to jump the gun, but I don't see that happening without some major Mono-fu or a working port to Java already coming down the pipe which are both a tad improbable.

-Adam

Adam Miller said...

Mether: I will admit that I don't know all the details but I would like to think that something that's as big of a deal as "Oh, to all our pure-linux shops out there you're screwed by our new product" would atleast merit some sort of a public relations outreach to the community with a clear roadmap or statement of some kind as to their reasoning.

-Adam

Jef Spaleta said...

Adam,

What is the communicated release date for the windows-only product that you are referring to?

The link you reference doesn't put a date on it. Are you sure the announcement made at the summit included a release date for the windows-only product?

Adam Miller said...

Jef: There isn't currently a release date, but as I have a RHEV channel available in RHN right now (while empty at the moment) I am willing to venture to say that it's right around the corner if the infrastructure is in place to distribute it.

-Adam

Jef Spaleta said...

So you've moved on from rumor into speculation? This is going downhill. Instead of clarity you are introducing additional assumptions into the conversation. I think you are setting yourself up to be angry regardless of the verifiable facts..and that's doesn't make for a healthy or constructive conversation.

On the page you referenced... there are 6 listed virtualization products listed at the bottom. Three of are available now as part of the newly released RHEL5.4 Three of which are listed as coming soon. Are you sure you aren't misconstruing the announcement concerning functionality available in RHEL5.4 as including more than it does?

-jef

Unknown said...

Sure, it would be much better to spend a year rewriting the management software to Java and then be late to the virt party...

...NOT!

The fact is that if you fail to bring the product to the market asap, you'll miss the boat and you ain't getting a second chance.

And also, the customers don't really care, 99% of them live in a heterogenous environment, not a linux-only bubble.

Jef Spaleta said...

jakub:

It still important for customers willing to pay for a cross platform solution to speak up.

There is a more interesting discussion here however....

If the codebase in question is really writing in .Net that is a very interesting position that Red Hat has found itself in. Customer interest in a cross platform solution may drive Red Hat into embracing mono as a near term technology solution instead of taking the time to port the codebase to java completely. It really all depends on how motivated paying customers are in communicating their desire for a cross-platform solution and how difficult it is to do the port.

-jef

mether said...

Adam

"Pure" Linux shops are extremely rare. Almost every customer has a Windows box around for one reason or the other. Waiting for years to release a product isn't a option either after an acquisition. I would be curious to hear your solution.

Sita-pati das said...

As others have pointed out, the RHEL adminstration piece is in production. Qumranet when they were acquired only had the Windows piece. Getting the thing to market was the first priority. Getting the RHEL administration piece is next.

No dark conspiracy to promote vendor lock in to a competing OS there... (as if!)

Unknown said...

I beleve what you said because another guy confirmed it.

I can tell you that the company I work for is developing a small but useful tool developed under a GPL licence and supporting KVM/XEN and a lot of features (Provisioning with Cobbler, puppet, and so on) : Symbolic.

www.opensymbolic.org


Take a look, and if you need help, just ask me ;-)

bye
Luca