This topic has some direct bearing on another recent post, the one regarding the future of Microsoft, but it's really its own story, which Microsoft just happens to own a small piece of.

The topic is virtualization, and the impetus behind the post is this excellent article at Information World on the subject and the major players in the market. VMware, of course, gets top billing, and as per usual, Microsoft features somewhere in the background as the looming villain.

The article posits a future in which all our needs are met virtually, a bit like the 1950's vision of a flying car that could whisk Dad to work downtown in the morning, allow Mom to make a quick trip for groceries in the evening, and fly the whole family up-state for the weekend getaway. Given the power of today's hardware and the relative inefficiency of traditional operating system design and utilization, this paints a fairly attractive prospect for many CIOs and responsible corporate officers. Virtualizing a number of specialized server systems on a single piece of hardware presents, as we all know by now, considerable cost savings. But until recently, you were still losing performance in most virtualized environments due to the need to run a base OS with the virtualization software atop it-a sort of silver thread tying the virtualized utopia back to the dark and dreary real world.

Now, the advent of hypervisor management layers is rendering that base OS obsolete-or so VMware would like you to believe. And this could have a profound impact on Microsoft's market-share in the server market and make many of their core competencies more or less irrelevant.

Microsoft is releasing its own answer in the hypervisor battle but as IW points out they are, as usual, well behind schedule with it. And it's a fair bet that in their incessant need to tie everything back to the legacy Windows code base, they have included much that is irrelevant in the hypervisor realm-just as their existing Windows Server software packs in a great deal of cruft that has no place on the average server. In the server market, this means on a performance basis, given equal hardware, Windows is frequently blown out of the water by Linux and other stripped-down operating systems. This has become less and less relevant as hardware has become more and more overpowered... but the advent of widespread virtualization is a step into the past when it comes to hardware, allowing companies to demand maximum performance from their machines 24/7, and price per performance is again becoming a factor.

If this vaguely circular theme sounds familiar to you, it should; it's a common one in IT, whether you are looking at the similarities between the big iron server data processing model of the seventies and eightes and the client-server model of modern web-based SaaS. And in this case, the comparison to look at comes from the early days of the personal computer, when many applications were essentially coded to interact directly with the PC hardware-becoming their own operating systems by default, because there was no operating system so efficient as to be able to host them at reasonable speeds on the hardware available in the day. As with the hypervisor model, the approach demanded maximum efficiency from the hardware, and a separate operating system for a specialized application only got in the way.

So one must ask oneself, why did computing evolve away from that in the first place? In part, as I have already alluded, it was simply the brute force that became available with advances in hardware design and manufacture-you could, at some point, afford to run Windows underneath your applications because there was enough juice to do so without slowing you to a standstill (although that moment came well after the initial releases of Windows and Mac OS! You'll also note that I am discounting DOS; a lot of applications did run on DOS, of course, but DOS itself was hardly an operating system in the modern sense of the term-Disk Operating System was pretty much exactly what it said, a clever widget to allow standardized interface with a disk).

But partly it was because it became more useful to run various applications in the same operating environment, because the interaction between them, whether direct or simply by virtue of being in front of your eyes simultaneously. Swapping disks to gain performance in specialized applications was not, on the whole, as efficient as running those applications less swiftly but more cohesively. Standards of interface and interaction were the ultimate efficiency, as it happened.

Now, however, interface and interaction are increasingly being defined at the level of the network, rather than the individual server or PC. Running applications simultaneously on a single operating system instance is no longer key. And we are back to gaining efficiency by maximizing utilization of the base hardware.

It's an open question how quickly the market will get this, or even if it will agree with this assessment in the first place. But if it does then VMware and other pure hypervisor proponents are likely to win where Microsoft and other heavy server OS software providers will lose.