The End of Xen?
Brian Madden has a fascinating
post on his site. The title pretty much tells you his central theses: "Prediction: Citrix will drop the open source Xen hypervisor for Hyper-V. The rest of the open source world drops Xen for KVM."
His main ideas are that XenServer, Citrix's hypervisor, will die off from lacket of market penetration. (Update: I misinterpreted Brian's remarks. He was referring to the open source Xen hypervisor, not Citrix' commercial XenServer implementation). Citrix will start using Hyper-V, from good virtualization buddy Microsoft. At the same time, Brian predicts that the open source community will rally around "KVM," which stands for "Kernel Virtual Machine." KVM has been integrated into the Linux kernel and provides virtualization services, so every Linux server now has a built-in hypervisor. This is further bad news for Xen, Brian argues, as the open source community will abandon Xen development and throw its muscle behind KVM development instead.
Whew! That's a lot of predictions. And they're well reasoned. Will they come true? That's another matter entirely. It seems to me that Brian is a little too dismissive of Citrix's commitment to XenServer. (It's not the first time I've thought that.) I don't see the rationale in spending $500 million to buy a company, only to give up on development of its chief product, which has now become one of your chief products, after less than a year. They're more forward-thinking than that, in my opinion.
(Update: These arguments, although made about XenServer, also apply to Xen. Citrix has sunk a substantial amount of time and money in the Xen community, and all indications to this point show that commitment continuing.)
Additionally, I see no indications, none, that development has slowed around XenServer or the other products appearing in the Xen line, like XenApp, XenDesktop, etc. XenServer is the core of that; would Citrix build those products to work on Hyper-V rather than Xen? Not until there's a lot more market information available that XenServer isn't selling, and won't sell in the future.
I also disagree somewhat on Brian's assertion that
"Hyper-V and the open source Xen hypervisor are so similar, in fact, that one could plausibly argue that Hyper-V is the "Windows version of Xen.""
I'm not sure what Brian's sources are on that, but I've talked to people in the know for both Microsoft and Citrix, and they state that although the two hypervisors interoperate very well, that they are not duplicates, or near duplicates, of each other. They were developed entirely separately, but there is a common perception, in fact, that Hyper-V is based upon Xen. Not true. (Note that I'm not saying Brian is implying this, because he's not. But I do take issue with the general belief that Hyper-V is a Microsoft photocopy of Xen.)
Brian then builds the rest of his argument about the possible ascendency of KVM on the foundation that Citrix will abandon Xen. Certainly, one could see that happening, if Citrix pulls the plug on XenServer. It's also no minor announcement that Red Hat, a couple of weeks ago, announced that it chose KVM over Xen in its embedded hypervisor. The bulk of other vendors, like Virtual Iron, Sun, Novell and Oracle are still building on Xen, though, so I'm not confident that this is the beginning of a rush to KVM and the end of Xen as we know it.
Brian brings up many interesting points, but in the end, I think he's a bit premature. What do you think?
Posted by Keith Ward on 06/30/2008 at 12:48 PM