Open Source Virtualization: No Reason To Celebrate

Sep 21, 2011 (06:09 AM EDT)

Read the Original Article at

The Open Virtualization Alliance is celebrating its six-month mark, having reached 200 company members, up from the 10 that founded the group in May to tout the existence of an open source, Linux- based option for virtualizing servers. But what is there to celebrate besides membership numbers? Proprietary virtualization from VMware, Microsoft, and Citrix Systems became more deeply entrenched than ever over these last six months.

The alliance is supposed to spread the word that an open source alternative exists to the proprietary code. The OVA will document the KVM system, which is a highly efficient virtual machine hypervisor built into the Linux kernel. It also plans to aggregate and publicize best practices, supply deployment expertise, and air successful implementations, all in a bid to increase the popularity of the open source hypervisor--and slow down the VMware juggernaut.

Hewlett-Packard, IBM, Intel, AMD, Red Hat, SUSE, BMC, and CA Technologies are examples of the muscle supporting the alliance. As a matter of fact, the first five used to be big backers of the open source Xen hypervisor and Xen development project. Throw in the fact Novell was an early backer of Xen as the owner of SUSE, and you have six of the same suspects. What happened to support for Xen? For one, the company behind the project, XenSource, got acquired by Citrix. That took Xen out of the strictly open source camp and moved it several steps closer to the Microsoft camp, since Citrix and Microsoft have been close partners for over 20 years.

[It's not generally know that VMware holds seven of the top 17 SPECvirt benchmarks for virtual machine operation. KVM holds the other ten.]

Xen is still open source code, but its backers found reasons (faster than you can say vMotion) to move on. The Open Virtualization Alliance still shares one thing in common with the Xen open source project. Both groups wish to slow VMware's rapid advance.

Xen was once the contender that IBM and others threw their weight behind as the most competitive candidate. How could pricey VMware compete with free open source? But to continue to appeal to open source developers, Xen needed to show its independence. Instead, soon after XenSource was acquired by Citrix, Citrix and Microsoft announced they would both support Microsoft's Virtual Hard Disk virtualization file format.

When KVM appeared, touting compatibility with the Linux kernel and high performance, a transfer of loyalties followed. Open source developers want the most independent and highest performing product available. KVM as part of the Linux kernel was winning benchmarks. Red Hat acquired KVM and encouraged its further development. Seeing the writing on the wall, IBM, HP, et al, migrated in KVM's direction. Better to switch horses in the middle of the race than watch VMware end up owning the racetrack, the grandstand, and all the stables out back.

IBM was particularly vocal in support of KVM, even when it wasn't clear the open source hypervisor had legs. Two months after the announcement of OVA, Jean Staten Healy, IBM's director of worldwide Linux, compared KVM to Linux in an interview with Database Trends and Applications: "Just like with Linux, open virtualization offers choice, lower costs, and interoperability. Those three elements are critical to businesses of all sizes and are the biggest reason why open virtualization is seeing such rapid and widespread adoption. In fact, you will probably hear me say those words a lot during our discussion because they really are key to understanding the need for KVM."

KVM is not like Linux, at least, not yet. Linux is growing fast, the only server operating system that's maintaining a 20% growth rate year over year. KVM usage is increasing but it still barely disturbs the scales when it's added in the presence of VMware, Microsoft, and Citrix.

Saying that KVM is like Linux, to me, is a non sequitur. If KVM is successful, it will not be because it's like Linux but because it is a well-designed part of Linux. It is both a value-add to and co-dependent with the operating system. Just because it is part of Linux doesn't mean that it's going to get used; if it maintains a performance advantage, on the other hand, and builds out a management environment, it may find itself being used where Linux is--in the enterprise data center and in the cloud. Although part of Linux, KVM can run virtual machines using any x86 operating system, including Windows.

The reason to celebrate the Open Virtualization Alliance is because it's throwing a spotlight on a dark horse option, far back in the race, but a horse with great potential. For greater enterprise acceptance, companies need to see what KVM can do, have its features and characteristics documented with stories of successful implementations, if they exist. KVM still lacks street cred among many users, who have been preoccupied with the Microsoft vs. VMware question.

If KVM is legitimately in the race, users of server virtualization are less likely to end up locked in to a single vendor environment. Some migration capabilities already exist among hypervisors, but KVM's existence as open source code makes it more likely that migration capabilities will increase. It makes it more likely that transfers from one virtual machine format to another should one day be routine, along with the ability to use multiple public clouds instead of being restricted to a single type.

KVM still lacks a complete management environment, but the growing OVA alliance says third parties, such as VKernel, are lining to supply pieces of it. VMware is threatening to run off with operations management in the virtualized part of the data center; IBM, HP, BMC and CA Technologies, the traditional systems management vendors, are all part of the OVA alliance and can be assumed eager to bring competing management capabilities to KVM.

VMware is setting a high standard for the managed, virtual machine environment. Microsoft and Citrix are racing to match that standard. But there's room for still another player, one that offers a distinct and non-proprietary choice.

The Open Virtualization Alliance needs to learn from the experience of its Xen project predecessor. It needs as open and active a development community as possible. Xen struggled with the perception that it was a project too much dominated by the interests of large vendors. The same fate could befall KVM, and Red Hat will need to guard against it.

In the past, VMware has shown an adeptness at competing with open source code, coming up with free versions of ESX Server and low-cost ways to get started with its product line just as a new open source initiative appeared. But the fully virtualized environment is getting so complex and expensive that some space has emerged between proprietary options and open source.

KVM, as GPL code, exists as a viable alternative with its support from independent developers and its positioning inside Linux. If Red Hat and third parties show they are capable of supplying a management environment around it, KVM is going to get a significant push from the likes of Intel, BMC, CA Technologies, and HP. The latter is contributing code to improve the capabilities of monitoring and managing KVM virtual machines through HP's Insight Control systems management product. Likewise, Insight Control can feed information back into Microsoft Systems Center or Red Hat's virtualization management console or other management systems, allowing KVM to fit into a larger data center picture.

In this fashion, an open source option allows virtualization to contribute to the more efficient operation of the data center. Performance efficiency of virtual machines is a larger factor than it used to be. KVM appears to shine on performance and a competitive open source option helps keep a thriving and innovative segment of the industry on its best game.

Read our new report, State Of The IT Service Desk: Change Management Remains Key. Download the report now. (Free registration required.)