OK, let me put it this way: How do you think server consolidation is done? It’s not done by putting multiple applications on a server unless you’re a serious masochist. The best practices with MS systems (and in fact I do this with UNIX systems as well) is to put one application per server OS. Note that I made that point; per OS. The simple fact is that in modern business there’s HARDWARE consolidation going on thanks to virtualization. However, typically there is little consolidation of applications to a single host OS because of the fear of a single rogue app taking down multiple environments.

Ironically, due to the I/O constraints and slight performance hit of virtualization (not to mention the perceived “cheapness” of a non-hardware machine), this virtualization typically leads to server propagation. That is, where an application may once have used 4 servers, it may now use 5 (I see a 25% propagation for virtualized app environments where I work). This means that a good Windows admin will be in higher demand because there are more server INSTANCES than there would be if the “consolidation” had not taken place.

This is also completely ignoring where the perceived “cheapness” of the virtual machines leads to redundant application build-outs where none existed before. An application that was on one server sometimes gets two in a virtual environment because the amount of work of putting up a new instance is so small. The performance often has nothing to do with it because the application so under-utilized the hardware it was on originally that there’s sometimes an increase in performance going to the virtualized environment.

This is good for the industry in my opinion. We also run virtual Linux boxes hosted on our Virtual Infrastructure clusters (VMWare). Sure, Linux’ stability means we could run more than one app per instance… but why? Why not isolate the apps when you can throw up a virtual Linux box in about 10 minutes that has a kernel, basic user-mode tools and no GUI? Then load up your app and you’ve got a Linux box that can boot in 10 seconds flat and be rebooted without the users noticing an outage. Add in some redundancy on a second box and voila; you’ve just improved your end user’s experience, made your application enterprise-ready, but also added to the server propagation I mentioned earlier. This is almost inevitable.

I see no reason this is going to change any time soon. The FA listed the canceling of Network Admin programs at schools as being the reason they see this going away. I call poppycock on that, too. The reason these classes are dying is because they were set up in the late 90’s when “Network Admin” was a high-demand, high-pay career choice and suddenly everyone wanted to be in on the “wave of the future”. Well, reality hit, the bubble burst and we ended up with a job market flooded by people calling themselves “Network Admins” who could barely tell you how binary related to a subnet mask (that’s important to understand, by the way!) Now few people want to school in Network Administration, so these programs are getting canceled because of lack of demand. Now, it’s been a few years since the dot-com crash… so most of the paper MCSEs and CNEs have weeded themselves out of the market. Those few who remain are gunning for management… and power to ’em.