Didn't Red Hat Used To Love Xen?

KVM, Kernel Virtual Machine, got wrapped into the Linux kernel 2.6.20 back in October 2006. Now Red Hat is promoting KVM in Fedora. Should you care?Red Hat made a pretty big deal of Xen integration when they pushed v5 of their Enterprise Linux build. SUSE went with Xen. Ubuntu went with KVM. Now Red Hat is pushing KVM as the next big thing; you'll find it in Fedora builds. A beta is available for download. I guess if KVM ends up in all flavors of Linux, theoretically every Linux kernel could function as a KVM hypervisor instead of running some flavor of Xen. Will this make management and roll out of virtualization easier at some undetermined point in the future? Most likely.

Does the move increase customer confusion and muddle the open source virt waters? Yup. Will it turn heads away from Hyper-V or VMware server? Not likely.

For background, startup Quamranet has been working on server-based KVM to deliver desktop virt; the company specs 20 Solid Ice guests out of a four-way, 16-GB host, assuming users aren't running graphics-intensive apps. In addition to selling desktop virt, Quamranet also happens to be the "proud sponsor, maintainer and catalyst behind the open-source KVM project."

Click here for a white paper on KVM tech.

Back to Red Hat's Summit announcement last month. I'm not sure I agree with the company's pitch: Unlike many virtualization solutions in the market today, Red Hat's technologies enable customers to deploy a virtualization infrastructure that is flexible enough to meet their individual business needs (from the 6/18 press release.)

I'm not contesting the flexibility of Red Hat's goods, nor the tech behind KVM.

I am a touch critical of the marketing digs against the competition. Is Red Hat being critical of Xen again? The company changed its tune once it got on the Xen train with Enterprise 5.0. Is the quip a dig against VMware or Hyper-V? Perhaps Parallels, Sun, or Oracle? Or any other smaller virt players? If so, it doesn't really ring true.

Off-the-record feedback from Xen community members at last month's Xen Summit hints that Red Hat hasn't been the strongest Xen partner. Nor the most cooperative over time. But that's another post.

Perhaps Red Hat has been a bit short on engineering resources for VM work and decided KVM is a simpler way to go. Perhaps Red Hat's management believes KVM is a better end-state technical solution for virtualization.

Either way, I'd be a touch anxious if I had a large Red Hat Xen install.

I caught up with Simon Crosby, CTO of Citrix and Xen evangelist, at the end of last week to get his two cents on Red Hat's KVM play. While he didn't use the phrase "flip flop," Crosby noted Red Hat's somewhat inconsistent virtualization strategy over time, and made the point that the virt market is moving beyond the hypervisor. While KVM gives one-stop-shopping for Linux build + hypervisor, customers are looking for scalable end-to-end solutions for robust VM management and application delivery. Yes, KVM can leverage native Linux management tools to make life easier for Linux admins, but native and third-party management tools for VMware, Microsoft, and Xen-based solutions are in a healthy features race, with management of heterogeneous environments a growing customer requirement.

As cliche as it sounds, virtualization is now being driven by business needs; customers want robust solutions with flexibility, enterprise management, and performance tuning across large, multihost installations.

Is KVM a "better" tech solution for virt on Linux? Again, likely. Will that make a difference in enterprise purchasing decisions? Do admins really mind managing an additional, feature-rich, robust hypervisor platform layer that works? I don't think so, and I don't see Red Hat getting many "switchers" based on KVM.KVM, Kernel Virtual Machine, got wrapped into the Linux kernel 2.6.20 back in October, 2006. Now Red Hat is promoting KVM in Fedora. Should you care?