Initial Reaction: Red Hat Enterprise Virtualization for Servers
I haven't seen it mentioned on Slashdot or LWN yet... and I even emailed LWN informing them... but Red Hat released Red Hat Enterprise Virtualization for Servers and the accompanying Red Hat Enterprise Virtualization Hypervisor products today. There are a slew of press releases and demo videos. They even had a webcast press conference. Oh, and hey, they also released all of the manuals too.
Didn't we see this coming?
Those who have been paying attention were expecting these releases... but I must admit that once I learned the details, I was shocked. Why? Because the Management side of the product requires Microsoft Windows products. What products? Windows Server 2003 is needed to run the management server. What services is the Windows Server running? IIS, Microsoft SQL Server, and it'll need to be connected to a Microsoft Active Directory Server for authentication and management.
Then of course once you have the proper Windows environment established you can actually install the RHELfS management app provided by Red Hat. Oh, if you don't have your own Microsoft SQL Server setup already, they can install the Express edition for you.
The Management server software includes a local management client app you can run on the management server but it also has a web-based version of the management app for remote management. Yeah, at last open standards! Not so fast. What browsers are supported by the web-based management app? Microsoft Internet Explorer 6 and up. Other browsers (according to the docs) are "untested". What? They haven't even tested them? What, you aren't curious if Firefox works? To clarify it appears as if the web-based management app sends Active-X based content to the web client... and that .Net technology is also used.
Ok, after reading those last few paragraphs, I'm guessing you are shocked too.
Everyone likes pragmatism, right?
In Red Hat's defense you have to realize that the product that this is based on was originally written by Qumranet before Red Hat took them over. The design was so dense and stacked in the Windows environment that it is going to take some hard core development effort to get it ported to Linux. The question is, is Red Hat currently interested in doing that? They have given some indicators that they are... but one has to wonder how compatible that idea is with the launch of the product line today. The setup has a zillion features and compares to VMware's complete product line nicely... but just how easy is it is going to be to later introduce a Linux version of the system and then having to maintain and support both?
Oh, almost forgot to mention that the Hypervisor part of the equation... that is to say the physical machines the management system is managing... are all Linux based. The Hypervisor machines can be either the very thin Red Hat Enterprise Virtualization Hypervisor (that comes with RHEVfS) OR machines running RHEL 5.4. At last, some Linux!
How big of an issue is all of this Windows stuff going to be? How disappointed is the grander Linux community going to be? Will there be any fallout as a result? I'm hoping it isn't going to be a big issue. I'm hoping that people won't be as disappointed as I was initially... and I hope there is no fallout. I hope that Red Hat is able to get the management system ported to Linux at some point... but if not, I'll get over it.
Some things I'm curious about...
How long did it take Qumranet to develop the Management system on Windows? How does the new release today compare to the SolidICE product that existed before the Red Hat merger? How many features had to be added to the new product that weren't in the SolidICE product? Is Linux missing any software components that will be a barrier to porting the management system to Linux?
I'm sure it wasn't an easy decision for Red Hat. I'm sure everyone involved had to closely examine the whole, "...but the management server is based on Microsoft Windows," concept and carefully weigh the pros and cons. I'm sure there wasn't much doubting of the functionality and feature set of the product... but more about the politics. If you were Red Hat, what would you have done? No, no... really.
But wait, there's some good news too
VMware really has a monopoly in the virtualization market. They started over 10 years ago.
Then along came Xen. It was free software but it had the problem of not having been developed the way the Linux mainline kernel developers prefer. Xen was developed and matured for years outside of the mainline development process and is quite an intrusive thing to try to get into the kernel these days... and hey, the hypervisor is really a different non-Linux OS, right?
Then along came KVM. It takes the hypervisor concept and implements it as addon modules to the Linux kernel... and uses all of the power that Linux already comes with... and doesn't have the duplication of code a hypervisor that is a different OS has. KVM is the clear winner... although I don't exactly count the Xen guys out yet... because continuing competition is good.
Oh, yeah... what is the good news? Well, Red Hat had virtualization before today but they didn't really market it as a real competitor to VMware's enterprise products. With the release of the new product line today by Red Hat, all of that has changed. VMware now has some competition. Look out VMware, Red Hat just painted a big target you. It had to happen someday, right?
Getting in bed with the devil?
People want to virtualize Windows on Linux. They just do. Oh sure, there will be some that only virtualize Linux on Linux (and you should know I'm a big OpenVZ fan that is for Linux-on-Linux only)... but people want to be able to do both. When the RHEV for Desktops comes out, how much do you want to bet that the majority of desktops that get virtualized are Windows? I'll take the bet and I'm not a gambling man. The fact is that Windows is popular and for Red Hat to compete it has to be in the equation. I was just hoping that it didn't play as big of a role in the Management system as it does... but I'll get over it.
Where to go from here
I watched the webcast. I read the info sheet and the press releases. I watched the demo videos. I guess I'll continue on with the manuals... as there are ~400 pages of documentation I've printed to wade through. I've already made a healthy dent in it.
Pricing you ask? Well, there is only one product and it includes everything... but there are two levels of support and you pay for the level you want as a subscription. It's $499 per socket per year for the Standard support. That's $998 for a two socket machine. The processors in those sockets could be dual core, quad core, or i7 (I forget how many cores the i7 has). While that may sound expensive to some, anyone familiar with the myriad of pieces and prices that are the VMware offerings can tell you Red Hat's product is simple (one product with all the pieces) and cheaper than VMware's. In fact if you total up all of VMware's various addons it can come out to as much as $3,000 per socket.
What remains to be seen is if Red Hat's product lives up to the feature set it has. Can it do well everything it claims it can? Remember, this is a first release. That isn't quite true considering the Qumranet products that came before... but so far as Red Hat goes, it is a completely new ball game.
I've been talking to their PR folks for a couple of weeks now trying to secure an interview... and I gave them my question list some time ago... but I'm guessing they were delaying it because many of the answers were just around the corner with the official product announcement. Now that I've seen everything most of my questions are no longer needed... and I've come up with several more. Once I get done processing all of the info, I'll redo my questions and hopefully have an interview article in the not too distant future... assuming I didn't offend someone with this blog posting. :) Wish me luck!