Stewart Alsop's latest column at Fortune talks about the virtualization of IT resources, that is, making the location and provisioning and management of processing transparent to IT personnel and end users. Think of this as a pragmatic interpretation of 'grid computing' or IBM's much hyped autonomic computing direction. Stewart proposes four immediate impacts: server consolidation, the death of appliances, reemergence of ASPs, and web services tools. While I fundamentally agree with the basic trend that Stewart is describing (anyone who's been near computer operations will immediately understand the value), I take some issue with the specific impacts described.

When I first skimmed the heading 'server consolidation' I read it as platform consolidation, and was then taken aback to find it projected instead physical consolidation of the server equipment. I'll stick with the original reading as being the impact. While in theory one might 'virtualize' with a covering API any collection of operating system and services architecture, in practice, programming, provisioning, and management are going to be a lot easier if the underlying substrate is uniform. In short, virtualization reemphasizes already existing network effects that are pushing toward consolidation of the server O/S and application platform. Who are winners and losers? Surprisingly, the clear winner is Linux. With IBM solidly behind it, and wide support in all corners (save one) of the web services movement, it's here to stay. The probably second survivor is Microsoft, simply due to its market presence. However, virtualization is going to make the strategic baggage that MSFT brings to the applications layers of the web services battle an impediment. The probable loser is Sun. When the dust settles, there won't be much room for Solaris, shackled to a high cost hardware base.

Jumping to web services tools, these will indeed be highly important as the enablers for virtualization. I'm dubious, however, that they will be a broadly investable area. Tools are likely to turn into a sort of scorched earth territory, with the warring parties in the server struggle discounting so heavily that extracting revenue and getting an exit will just as hard as - well - programming tools have always been.

The death of appliances, physical consolidation of the server, and renewal of the ASP model are all questions of where scale economies will appear. I'll sign on for the appliances forecast. They have been a way of simplifying the install and management process for new functionality as it appears (and not coincidentally simplifying the revenue model for their vendors). As virtualization spreads, the install becomes hardware independent. I'll make an exception for 'appliances' that are largely about management of IP traffic - they will remain.

Physical server consolidation and ASP revival are far from certain from my point of view. With a virtualized (or uniform) server substrate, it becomes feasible to manage server farms like Google's 20,000 machine cluster. Even for IT shops with more diverse application bases, it will allow an independent choice of software and implementing hardware. The outcome is a complex conversation among the optimal mix of form factor and processor power in a changing Moore's Law environment. The removal of functionality outside the firewall to ASPs would still seem to be played out on application by application basis. If nothing else, the natural resistence of an IT staff to have its own expertise gutted by moving large chunks of functionality outside will put a damper on such a movement, at least at larger companies.

For another interesting view on virtualization, see this comment at VentureBlog.
9:45:06 AM