News

Where Virtualization and Open Source Collide

• By Ed Scannell and Doug Barney
• 03/01/2008

Citrix's Levine believes Microsoft will cooperate in helping set virtualization standards.

The next few years should be interesting times for Peter Levine, senior vice president and general manager of Citrix Systems Inc.'s Virtualization and Management Division. One would hope it's not the "interesting times" the Chinese talk about. Levine's job responsibility puts him at the intersection of where the worlds of the virtualization market and open source collide. On the virtualization side he runs into the white-hot VMware Inc., as well as the imposing threat posed by Microsoft. On the open source side he figures to knock heads with the leading lights in the Linux world.

Levine's resume makes him competitive, though. As president and CEO of XenSource, he helped turn the startup into a commanding presence in the open source world. He was in on the ground floor of Veritas Software, where he helped grow that company from zero revenues to $1.5 billion. For more than 11 years Levine also served as a managing director at the Mayfield Fund, where he invested in a number of software companies that became notable, including Mendocino Software, Centrify Corp. and Zenprise.

Levine sat down with Redmond Editor in Chief Doug Barney and Editor Ed Scannell, as well as Keith Ward, editor of the upcoming Virtualization Review, to discuss a number of topics ranging from balancing Citrix's Windows and open source development strategies to, of course, balancing its complex relationship with Microsoft.

Redmond: How do you see the Citrix-XenSource relationship with Microsoft: as direct competitors or working to complement each other's platforms?
Levine: One of the reasons Citrix and XenSource came together was because of our working relationship with Microsoft. Each company has a good relationship [with Microsoft] that has endured for years. As for being direct competitors or more complementary, there are features we have that can be brought over to Viridian [now Hyper-V]. I'll expect us to have two product areas based on two different hypervisors.

When approaching corporate customers, in which cases would you promote the Xen core platforms, and in which your services running on top of Hyper-V?
We're about building a virtualization platform that runs without any OS agenda, so we'll run a bare-metal virtualization. In the case of Microsoft, where there's more of an OS agenda, some folks may want to first adopt the Microsoft platform and then build services on top of that. In other cases, some may want to build bare-metal virtualization and then build services on top of that. I also think there will be a class of customer who wants the latest and greatest virtualization features, in which case we'll be able to offer those on top of Xen.

Windows Server 2008 has a lot of virtualization capabilities built in. What will be the overall impact on the virtualization landscape of Windows Server 2008?
We have to see just how much virtualization technology will be in Windows Server 2008. You must have a certain level of features for people to consider using the technology. Assuming Microsoft achieves that base-level functionality, we'll be happy to deploy either Viridian or Xen as the base platform. We'll build our value-added services on top of either of those baselines. Remember, Xen is open source, so we've already given it away. We don't derive any revenues from that engine the same way we do from the Viridian engine. All of our revenues are derived from building the dynamic virtualization services on top of the underlying hypervisor. So to the extent that Microsoft is successful with its virtualization platform, we hope to be successful with the add-on business we build on top of it.

Were you surprised when you saw the initial pricing of Windows Server 2008? Specifically, that Hyper-V is going to sell as a separate product for $28 total.
Well, I'm coming from the open source space where the Xen engine is free. So it isn't $28 or $10, it's $0. To have broad adoption of an underlying base, you want to drive the commodity components to zero and then build the added-value components for revenues. The notion of having hypervisors be ubiquitous and having the value-added services be the things that users actually pay for is perfectly consistent with what we've done at Citrix-Xen, and perfectly consistent with our model of building on top of the Viridian base.

What are the challenges you need to overcome to get the laptop engaged in this new world of virtualization?
We need to understand what the appropriate use cases are for those types of environments. One is just figuring out what the customer absolutely wants. There's the challenge of a virtualized environment out on the laptop and how that actually all happens-from a business standpoint, where does that actually get done. Then, what are all the infrastructure pieces between the back-end server and that laptop to make it all work correctly? The challenge is less of a technical issue than it is a use case or business issue.

There are several companies with full solutions that have virtualization platforms. How is this going to play out in the marketplace among VMware, Citrix and Microsoft?
We have to separate the platform from the added-value services. One is the virtualization stack and the other is the platform. Citrix does both. I think a lot of that platform exposure, whether it's VMware, Citrix-Xen or Microsoft, is going to be driven by the hardware vendors who actually put these servers out there. A big part of the not-too-distant future is making the platform a property of the hardware itself. If a company is blessed enough to have the virtualization platform working with the hardware at the time of purchase, that simplifies the entire user experience.

What kind of trouble do you anticipate Microsoft will be in terms of setting standards for a virtualization stack?
We'll have to wait and see. We have a very close relationship with Microsoft; they seem to show a willingness to have open interfaces.

Are you working through any industry bodies now to establish standards? If so, which ones?
We announced an open standard in September [2007] that VMware and Microsoft are supporting. This guarantees interoperability in the underlying platform, so that's a start. The real openness is not so much a Windows guest that runs on top of XenSource or runs on top of VMware. That's trivial. It's the tools and services that get built around it. This is where we'd like to see more collaboration and innovation.

Given the rising price performance of multi-core based servers, many prefer to just buy more servers instead of going to a virtualization strategy. How much does this impede your efforts to sell virtualization products?
VMware has been successful on the server consolidation front, which was generation one. However, the notion of virtualization creates an agility and manageability in those environments, so that the true value of virtualization is more about those attributes of an environment as opposed to server consolidation. I can envision a world where every server is virtualized. Only 8 percent to 10 percent of the new server market is virtualized right now. The use cases for virtualization have only started to be well understood.

Some believe virtualization should fundamentally change how IT buys servers. The idea is you buy a large server beyond your needs, set it up as a virtual server, add applications until it fills up, then buy the next one. Is that something you're seeing?
That's an interesting use model. A big server is going to be much more expensive than a lot of little servers so the question is: How does it get traded off against power, cooling and manageability? One could argue that you could buy five blades today, and as your environment grows you add new blades. So it really becomes a space and power issue and, to some extent, a management issue.

I think virtualization gives you the agility and flexibility to either carve up a server into little pieces, or have lots of blade servers. It gives you the ability to have a dynamic infrastructure, whether it's a big server or lots of blade servers. I'd agree that people are starting to deploy systems in that fashion.