Server virtualisation involves making two or more “logical” servers on one physical server. That is, we have one “box” with one processor (though more are possible and will make things run better), one set of memory and one set of hard disks — but onto that, we build multiple instances of servers for various uses. Each of these servers is totally independent of the other and exists only in software — they are of course linked by the shared hardware. Each has its own name and network address, and can be rebooted without affecting the others on the same box.
Typically, these servers could be used for any function, but the specifications of the server must grow in accordance with their use. For example, a box with virtual servers which perform the functions of active directory, DNS, a small web server and printing will need fewer processors, memory and disks than one with a couple of databases running the corporate intranet and document management functions.
So what’s it for? If space and budgets are tight, it is useful to make the best use of hardware. Practices tend to deploy single applications to single servers. In the first example above, I would require four actual servers, but in the virtual world, only one. The magic lies in the fact that most servers run at around 5-10% capacity most of the time.
Virtualisation makes best use of this by consolidating many servers into one — some typical ratios are 10:1 and 15:1. This could save you a lot of time and money.
Virtualisation is also very quick if you need to roll out new servers with new applications or websites. Traditionally, you might have bought a new server, waited for delivery, constructed it and racked it into the cabinet, put Windows on and patched it. With virtualisation, you can simply build a new virtual server on an existing physical server and be up and running in hours rather than days or weeks.
Downsides? It is important not to overload the physical server with too many virtual servers and swamp its resources. The physical server you use also needs some redundant components — hard disks, power supplies, fans.
If you need to power down the box to make a change to one of these items, you will now be affecting many servers and functions rather than just one.
Cost, resource efficiency and speed of provisioning are the key drivers, although the price you might pay is having all your eggs in the same basket.