News

Taking Stock: Microsoft's Virtual Power Play

• By Stephen Swoyer
• July 30, 2008

Just how favorably does Hyper-V compare to established products from VMware Inc., Virtual Iron Inc. and Citrix Systems Inc., to name a few? How serious is Microsoft about competing in the virtual space?

Both are complicated questions. When it comes to virtualization, Microsoft isn't exactly viewed as a thought leader. There's an inescapable sense, after all, in which virtualization -- because it abstracts both operating system and application assets -- actually displaces the OS. Given Redmond's OS-centric business model, a disruption of this kind seems like a clear conflict.

That might account for Microsoft's protracted hedging of its bets with respect to virtualization: It was against it, in other words, before it was for it; it accommodated it, but it also wanted to pigeonhole it -- as anterior to the OS itself. By making several virtualization-specific acquisitions, Microsoft spent a sizeable amount of money precisely to pigeonhole it.

Late last year, Redmond bit the proverbial bullet, announcing an upcoming hypervisor offering (the eventual Hyper-V) that -- to a surprising degree -- looked a lot like everyone else's.

Virtual Disruption
Redmond had plenty to be afraid of. Virtualization technologies are disruptive because they're more fundamental than the OS. In the virtual space, the OS -- the former bedrock of one's application assets -- becomes just another asset or resource.

Microsoft, not surprisingly, initially viewed this as a threat to the dominance of its Windows brand. For the better part of 48 months, it struggled to embrace -- and co-opt -- virtualization on its own, Windows-friendly terms. Even though competitors VMware, Virtual Iron and Citrix were selling hypervisors that functioned independently of any specific operating environment, Microsoft seemed determined to push an operating system (i.e., Windows Server 2008) that is the hypervisor.

That didn't work out so well, so Redmond adjusted course last November, and announced a standalone hypervisor -- its Hyper-V Server (derived from Hyper-V, the official brand for the former Viridian technology) -- that would allow customers to virtualize heterogeneous workloads onto a single server.

In this regard, experts said, Hyper-V isn't all that different from VMware's ESX Server or any of a dozen other prominent hypervisors.

"Microsoft's decision to offer a hypervisor that's not part of the operating system [is] striking, given that they have been the most vocal proponents of the 'virtualization-as-a-feature-of-the-OS' point of view," said industry veteran Gordon Haff, a senior IT advisor with consultancy Illuminata, at the time.

Since then, Haff has had a lot of time to think about Microsoft's virtualization strategy. An idea can do a lot of gestating in nine months, and Haff -- who was initially high on Redmond's Hyper-V about-face -- sounds even more optimistic (and more pragmatic, too) these days.

That's in spite of the fact, Haff insisted, that Microsoft can't really compete with the big players, at least when it comes to features, performance and technology. Redmond's edge, he stressed, is the ubiquity of its Windows brand.

"If it were some arbitrary company, Microsoft's virtualization product portfolio wouldn't be mature and broad enough to deserve discussion in the same vein as Citrix and VMware," Haff said. "Microsoft has clearly been playing catch-up, especially in the hot area of server virtualization."

For a long time, according to Haff, Microsoft was content to trade blows with VMware, mainly to check that company's success and to blunt the potentially disruptive impact of virtualization itself. This meant accommodating virtualization on Microsoft's own terms, or -- in a now-familiar feat of triangulation -- "embracing and extending" that technology.

Redmond's acquisition of the former Connectix (with its still-gestating Virtual Server product) was an exercise in kind, according to Haff.

"[Virtual Server] gave Microsoft customers a server virtualization option that didn't involve heading over to VMware and thereby kept them more fully in Microsoft's fold. If Virtual Server wasn't up to the full virtual infrastructure play that VMware was increasingly running, it was sufficient for the basic slicing/dicing or test/dev provisioning that still represented how a lot of folks were using virtualization in practice," Haff said.

Virtual Server was far from a wash-out, he maintained. "If it ended up being largely a bridge for Microsoft customers while they waited for Microsoft to add integrated, native virtualization -- well, bridges can be awfully useful," he said.

Virtual Innovation?
It isn't as if Microsoft is playing catch-up with VMware and other competitors in vain. Haff cited Redmond's bleeding-edge take on application virtualization, which (admittedly) plays to its client-side strengths.

Application Virtualization -- called App-V -- is the fruit of Microsoft's 2006 acquisition of Boston-based Softricity (developer of Softgrid), so it has demonstrable best-of-breed roots. It also does away with one of the most onerous (and increasingly bloated) requirements of the Windows desktop model: the need for locally installed application instances.

"Applications are packaged and stored on a server, then 'streamed' down to the client system when requested. The application executes on the client using local hardware resources in the usual manner, but without the need to actually install it. Application virtualization isn't unique to Microsoft, but Microsoft's technology here is quite sophisticated," Haff said.

App-V, as implemented by Microsoft, is a Big Idea, inasmuch as it brings improved manageability, reliability and (putatively, at least) scalability to Windows applications. "This type of application virtualization also introduces the idea of application containers. These bundle together an application's files, along with other components, such as shared libraries and custom configuration settings, which are needed to run the application properly," Haff said.

That's just the tip of the iceberg, of course. "In addition to providing a management structure to keep all the necessary parts together, application containers also deal with some of the conflicts that can keep two applications -- or two versions of the same application -- from co-existing on a single system," he said. "A common problem is that applications require different versions of the same library [e.g., a Windows DLL file]. Containers encapsulate and isolate those libraries and applications in a way that lets them co-exist on a single system."

Softricity wasn't Microsoft's only App-V-related acquisition. As Haff noted, Redmond is still incorporating technology from related acquisitions, such as the former Kidaro and Calista.

Far From Virtual Authenticity
In light of its about-face with Hyper-V and its arguable innovation with App-V, Haff thinks Microsoft is dead-serious about virtualization. What's striking, he argues, is that Redmond now conceives of virtualization as a technology that can help it sell more Windows licenses, not fewer.

"Microsoft has largely gotten past these concerns or, at any rate, decided that they're manageable in a world that's heading pell-mell toward virtualization with or without it," he said. "Part of this was getting license plans in place that include the licensing of Windows guests as part of premium versions of Windows Server. In other words, Microsoft has discovered that virtualization makes a nice carrot to upsell Windows licenses."

Haff added, "As for virtual appliances, they're still used mostly for demo software; Microsoft is pushing application virtualization as an alternative to simplify software installation without forcing 'one size fits many' choices as the appliance model tends to."

So what kind of force will Microsoft amount to in an already teeming virtual space? A potent -- if not quite dominant -- one, according to Haff.

"If a company is largely a Windows shop today, and isn't looking to implement a particularly complicated virtualized infrastructure, Microsoft's products are probably the most natural, lowest-cost, best-integrated, easiest to acquire and install path to server virtualization," Haff said.

On the downside, Microsoft's VM management tooling won't ship until later this year, and other VM management scenarios such as "live migration" (i.e., the shifting of a running VM from one system to another) aren't yet as straightforward in Hyper-V as they are in VMware or Xen.

"Microsoft has said that adding such 'live migration' is a development priority," Haff said. "Also on deck are performance enhancements -- such as taking further advantage of on-chip virtualization accelerators -- and making VM memory use more dynamic."

Does this mean that Hyper-V is a laggard -- at least, compared to its established competitors? Not necessarily, Haff said: "It's not that Hyper-V is unusually incapable, but it's a first pass and its features, level of tuning, and consequent adoption rate will all reflect that."

What's most surprising, Haff said, is that virtualization -- far from harming Microsoft's bottom line -- seems to have boosted it. "For all the fuss that Microsoft has made about getting Hyper-V into play, this has never been as important as Microsoft thought it was. The reality is that Microsoft has sold lots of licenses -- operating system, application and middleware -- in and around VMware virtualization," he said.

"The bulk of VMware's business is in and around those same products. For all the animosity between the two companies, I see them as more complementary than competitive."