server side virtualization

Microsoft math drives Microsoft marketing. (Cue shareholder applause...).

But you don't need calculus to see why - despite public nods to the need for server-side application virtualization -- Microsoft will not be moving in that direction any time soon. Let's start with their April demo of a server side application virtualization product that is without name or estimated timeline for departure from the lab.

The use case shows a gold master application image running on a Windows 2003 server:

• The application is, in effect, "lifted up"
• The OS is upgraded from SV1 to SV2
• The application is returned to its place on top of the OS
• The application and the OS live happily ever after simplifying the life cycles of both OS and applications in the data center, saving money on numerous fronts, and bringing nice role clarity between system administration and application professionals.

First, let me point out that Microsoft has adopted the same approach to server-side application virtualization that AppZero pioneered - the decoupling of an application from the OS. Why? Because - for reasons I will gladly address in a follow-on blog - it's the only approach that is practical for movement across a network to any server (physical or virtual) in datacenters and across clouds.

Mary-Jo Foley noted in her April 14^th All About Microsoft zdnet post that Microsoft had pointed to an article that says, "If Microsoft releases Application Virtualization for Windows Server, it's not killing its Hyper-V strategy; it's implicitly suggesting to use hardware virtualization for OS delivery and application virtualization for services delivery." Well, there you have Microsoft singing AppZero's theme song.

Exactly.

That same post quotes a Microsoft Group Product Manager who said, "It (app virtualization for servers) is an area of interest to Microsoft customers. One of the key benefits is a significant reduction of server OS images. [emphasis mine] Instead of having an image for each instance of a server application, you'll just have one golden image for each OS you're using. If you need to patch that OS, you only do it to one image instead of all the images that have that OS."

My point.

Microsoft's model of one machine (physical or virtual) that runs --> one OS that runs --> one application with --> one set of application support infrastructure (anti-virus, system, and application management ...) forces the enterprise customer to buy a copy of a Windows OS for each application instance. The Microsoft yippee factor only goes up in the server virtualization world where virtual machines (VM) can be spun out so quickly that there are more OS instances than ever before. VM sprawl is an opportunity, not a problem in the Microsoft economy.

Not so with AppZero's current (and Microsoft's far-distant) approach to server application virtualization. The decoupling of application from OS effectively isolates applications from one another so that one OS can support 2, 3, 4, 5 ... applications. When you do the math, there's a big drop in the Microsoft yippee factor - with a corresponding surge in enterprise $$$$ savings.

Accelerating the marketing impact of this simple math is Microsoft's Q3 2009 earnings report in which only the Servers and Tools division posted significant revenue or income growth (7% and 24% respectively). By contrast, the Windows Client division saw revenue fall by 16% and income decline by 19% year-over-year. In a situation that eWeek called, "a stunning turnabout", the Server and Tools division's annuity-based revenue now exceeds that of the OEM-heavy, transactional Windows Client division.

License slashing server-side application virtualization technology from Microsoft? If you're really busy, call a third grader to do the math.