When people ask me which virtualization platform they should choose, I give them the same answer I offer to home users asking for PC or digital camera recommendations: "It depends."
Many people look at product reviews and research reports when choosing IT platforms. They want a quick "Consumer Reports" answer. Yet, when choosing core products, a critical step in evaluating vendors is examining their strategies and approaches, not specific products. So, in this comparison of three virtualization heavyweights, VMware, Citrix (which I'm pairing with Xen) and Microsoft, I'm looking at their positions more than their products At the end, we should understand the vendors' take on the virtualization market, its direction and the priorities chosen to achieve their goals.
This approach may fly in the face of product evaluation traditions; but it can be just as important as figuring out whether their products meet technical requirements. After all, if vendor X says that server virtualization is dead, and desktop virtualization is the future, buying vendor X for server virtualization would be a bad idea even if their product works well in that approach.
Citrix and XenSource
"We want a strong ecosystem. We want to enable choice."
Those are the words of Simon Crosby, CTO of Virtualization and Management Division at Citrix and former CTO of XenSource when asked about his top focus. Depending on where your own business priorities lie, this approach can be either
Historically, Citrix has had a solid history of supporting multiple ecosystems and encouraging their channel to support them. This approach has resulted in an impressive array of products being developed for their suite that address many verticals. ISVs invest in the Citrix machine because they see a viable business and – more importantly -- because they know that Citrix tries not to step on its friends. This is great for IT managers in these verticals.
From a product perspective, ecosystem building is nice but it doesn't tell you where a company is going with their engineers. Crosby stated his belief that hypervisors will be commoditized as core components of servers. He thinks virtualization has to be part of the "iron."
When virtualization is done right, said Crosby, it is part of the infrastructure itself. For example, in Crosby's world, the SAN (Storage Area Network) isn't virtualized by software; the SAN is instead aware of the virtual infrastructure and works optimally with virtual infrastructure. In some cases, this means helping infrastructure vendors by giving them the hypervisor. For others, it simply means running an open hypervisor platform.
In essence, Crosby is saying that the future of data centers is one where virtualization is so embedded in everything that needs to be done that the infrastructure knows how to abstract itself instead of depending on others to do the abstraction for it. The concept is definitely a little heady at first, but it doesn't take long to identify the potential of a fully-virtualized data center where everything is responsible for virtualizing itself.
As a result of this fully-commoditized vision, Citrix identifies their contribution at the application level. In short, how Citrix can deliver an application, regardless of what the application needs of its infrastructure, that makes both the operational and end user experience better.
This approach requires that everyone agree to play nice with virtual environments and standards to be established in this area. It's a tall order. Judging from the show of support from infrastructure vendors at Citrix's latest App Delivery Expo (iForum), however, Citrix's aggressive push is paying off.
The bottom line is Citrix/XenSource will differentiate based on what they do with their core technology and how they build a community a products around it to create the fully virtualized datacenter as opposed to differentiating based on the core technology itself. It is an atypical approach, but one that makes sense given XenSource's roots in open source software.
IT managers who agree with this philosophy, and it does have inherent value, have to determine if the ecosystem surrounding Citrix/XenSource is strong enough to stay the course. Those that do agreed should look at the Citrix/XenSource virtualization ecosystem and ask whether the products there help your applications run better. If the philosophy and the products fit, Citrix/XenSource should be on your final evaluation list.
"Virtualization is like TCP/IP. It's a means to an end."
David Greschler, Directory of Integrated Virtualization Strategy for Microsoft makes no bones about Microsoft's belief that hypervisor technology will become a commodity and uninteresting. Sounds like Microsoft and Citrix might be singing from an awfully similar hymnbook!
Historically, Microsoft reaches feature levels that satisfy most users such that many don't feel compelled to seek better alternatives, even when better alternatives exist. Right now, Microsoft isn't a strong alternative in virtualization. Greschler himself noted that "we aren't there yet," but he thinks Microsoft will be a strong, or even the strongest, alternative in the near future. In the meantime, someone looking at Microsoft's technology has to ask themselves whether today's implementation is indeed good enough to use until then, if it's better to wait and virtualize when Microsoft does deliver, or if the immediate need is so great that an alternative, like VMware, should be used right now.
In all its spheres, from the desktop to the server, tight integration is the focus and primary value of a Microsoft ecosystem. In virtualization, Microsoft promises depth of management, achieved with its System Center's integrated view of all the virtualization technologies used in a given data center. In virtualization, according to Greschler, consider than an application running inside of a SoftGrid application streaming virtual environment inside of a virtual server in the data center and piping its display using Terminal Services to a Virtual PC inside of someone's desktop has so many layers of encapsulation that traditional management tools simply don't have an accurate view of what's happening. Microsoft's focus with their System Center management suite is to provide that visibility down to the iron itself.
What's good about this? For one thing, the concept does address a challenge that happens with too many layers of encapsulation. Networking people have repeatedly experienced this phenomenon, most recently with encapsulating XML in HTTP, when the tools for monitoring no longer have sufficient depth to accurately see what's happening. In Microsoft's view of the virtualized data center, multiple layers of virtualization are inevitable. This, of course, means that tools to identify resource allocation and provisioning are going to be critical.
Is there anything that might not be so good? That all depends on whether you believe Microsoft's commoditized virtualization technology will deliver what you need for virtualization in the data center in the first place. Microsoft does have a strong position with their Virtual PC, Terminal Services and SoftGrid (Softricity) technologies. Hyper-V, once called Viridian, has yet to be seen.
The bottom line for Microsoft's position that the infrastructure doesn't matter. Focus on the management story. Operational teams may agree with this view, but data center architects may want to see some more out of the infrastructure before they look away.
"Being able to virtualize everything is the number one thing that drives VMware."
Raghu Raghuram, vice president of VMware Products and Solutions take on virtualization infrastructure offers a significant contrast to Microsoft's and Citrix's views. Where Microsoft and Citrix consider core virtualization technology as commodity, VMware sees the opposite: Core virtualization technology is the centerpiece. By focusing on what Raghuram calls the "Virtual Infrastructure", VMware looks to provide a fully virtualized data center where administrators see a single consistent view of everything in the data center even as the infrastructure itself changes.
VMware, in its relatively short history, has succeeded in virtualizing significant parts of the infrastructure -- such as SANs with VMFS -- and moving the ball forward with hypervisor technologies, such as live migration using VMotion. In other words, VMware's goal to fully abstract the data center environment so that all virtual hosts only see the view that VMware gives them appears to be an achievable endgame.
The significance of this approach is that it becomes possible to present a uniform view of the world, regardless of whether the world knows what to do with virtual machines or not. For data centers that would otherwise need to upgrade infrastructure to support virtualization, this can be a very appealing proposition; virtualization can be achieved without further investment. For system administrators and application owners, this means having a single view of the world around them, even when the physical world changes.
The challenge to this approach is that it remains a tall order to virtualize every conceivable piece of infrastructure and keep up with the changes in technology. History may be on VMware's side with respect to their ability to execute on this vision, but the challenge doesn't become any easier. If anything, success brings more work to virtualize more infrastructure.
For VMware, succeeding in this area would crack the holy grail of utility computing, where any unit of capacity is genuinely interchangeable with any other unit of capacity, regardless of the physical world that surrounds it.
From VMware's perspective, the bottom line for administrators is: Core technology is important to delivering what they see as a true potential for virtualizing the datacenter; that is, utility computing. It is here where you'll find the greatest strides forward by VMware.
VMware, Microsoft Virtual Server, or Xen? It depends
Between starting this article and reaching this point, I received no less than three "what's the best" questions about cameras, phones and laptops. I, in turn, frustrated no less than three buyers with my stock answer: "It depends."
Having heard the raw, unfiltered words of Citrix, Microsoft, and VMware and where they plan to put their research and development dollars I can firmly say that the same answer applies to, "What's the best virtualization solution?" It depends.
It depends on what you value the most in a solution. It depends on what you need to address in your own data center. It depends on what you need out of a virtualization technology today. It depends on what you need out of a virtualization technology tomorrow. It depends on what you expect from a management perspective. It depends on what you need out of vertically-focused virtualization solutions. It all depends.
Once you can succinctly define your own needs, re-read the positions of each vendor. Chances are that you'll notice yourself aligning with one over the rest. It's an approach that lacks the energy and fun that a pundit-fueled debate has or the simplicity of a "Consumer Reports" ranking. Also, it'll rob you of excuses to play Bejeweled during endless committee meetings.
The good news is that it will spare you from hearing smart-aleck, overpriced consults telling you: "It depends."
About the author: Steve Shah is principal at Rising Edge Consulting, based in the San Francisco area. A 15-year IT veteran, he is the author of the book, Linux Administration: A Beginners Guide, and a contributing author to books on content delivery networks, Unix and Red Hat Linux.
This was first published in December 2007