One of the things I think about from time to time is how appropriate it is to overbuy on resources either purely because I can or because I like whatever it is I'm buying.
People who aren't into computers tend to do this by buying large trucks, owning multiple cars, or an inordinate amount of whatever physical good goes along with their hobby. For some reason, it seems as though nobody thinks a second thought about this, but if you mention to somebody you own a large server, everybody asks or says things like "isn't that really expensive?", "I bet that takes a lot of electricity", or "what would you even do with that?"
This attitude is weird, maybe it's just that few people have actually shopped for servers, but in general, the answers are that it's actually not much more expensive than a commodity gaming computer, it takes less power than most gaming computers do, and it's very useful, especially for households or buildings with a lot of residents and individuals with a lot of data, especially given that desktops seem now to be less and less optimized for holding a large number of disks. I sometimes reply by asking how much it costs to buy a new big truck or how much it costs to insure them, especially compared to a compact or midsized sedan.
I think most people understand it when I make that comparison, but as always, I have my own doubts, wondering things like whether or not I have over-bought and what I should do for the replacement of any given machine.
Two cases where this is really true are superslab and TECT, which are certainly the most expensive machines I've personally purchased, and pretty close to the most expensive machines I've ever used. I consider superslab (a ThinkPad) to have been perfectly justified given the time in which I bought it, how long it has lasted (I'm still using it for a lot to this day) and what I've done with it in that time. TECT is a harder sell, because it has over-the-top expandability, but isn't that much faster than a desktop or single-processor server I could buy for half its cost. The question is how long I'll keep it and how far upgraded it will become over its life time. If I replace it in two or three years with something from a lower product tier, then there's a chance I have needlessly over-bought, but if I keep it for several more years after that and install a lot of upgrades, such as another processor and some more memory, then I'll know having bought such a large system was worthwhile.
But, even if I don't bring the system anywhere near its maximum potential, is it justifiable to have bought it? Can I simply use "I like it" as an excuse, or is it necessary to buy the biggest computer I can afford at any given time, even if the tasks it does can be completed by something much smaller? Or, if I'm sure the computing needs aren't that wild, should I be spending the money on additional ecosystem and infrastructure such as disk shelves or software?
One of the reasons this is kind of important is I do not know how much my data storage needs will change in the next several years, and arguably the component of TECT I am most fully utilizing is its storage system. I currently have eight 2TB disks, set up in such a way that gets me a 2TB boot/Exchange/SharePoint volume (mirrored on two disks) and a 10TB data partition (RAID 5 across six disks) and have been having good luck with this setup. It successfully holds all of the data I use on a regular basis, and a lot of stuff that formerly only survived in off-line formats such as on a single (very losable) flash disk or memory card is now on a redundant set of disks, on-line and searchable very quickly.
The next thing I need to think about, however, is protecting that data, and this is one of the areas where it always becomes pretty clear that over-buying these resources isn't always a good thing. It's one thing to say that you've got a big server, people seem to realize just how ridiculous the endeavor has become when you explain what's involved involved in truly protecting the large quantities of data worth saving that goes along with it. (Again, maybe this is only me.)
I've also looked into over-buying individual subsystems recently – nVidia has announced its high end "GeFORCE GTX Titan" card, a $1000 graphics card for individual desktop, enthusiast, and gaming PCs. By all measures of the word, it overkills, and it's approximately what I think of when I think about computing equivalents to the Ford F-150 Raptor. Is this a fair comparison? I think it is. Each of these things is over-priced for most people, and has functionality that really only has a limited application. In the case of the Titan, the Kelper architecture makes it essentially a miniaturized, slower, and cheaper Tesla card suitable for a reasonable amount of GPGPU computing. In the case of the Raptor, things like a larger and more durable engine, a better set of differentials, and downhill descent control make it better suited to its specific task, off-road travel.
I suppose part of the question is whether or not it can be considered over-buying if you can make complete use of any given system or tool, even if it's not on a regular basis. Take superslab the ThinkPad T400 for example. The other day, while trying out some photo workflow stuff, I am pretty sure I was using 100% of its subsystems at 100% or near capacity. Its processor and memory were busy both running a virtual machine and converting a "few" (twenty or so) gigabytes of photos from a camera RAW format to DNG, its USB and storage subsystems were busy with both things, its discrete graphics card was handling regular Windows 7 Aero tasks as well as the photo application and some web sites, and its networking was slightly less than 100% utilized, playing Pandora and accessing an SSH session.
Superslab isn't exactly new, I bought it more than four years ago at this point, and it's a laptop with laptop-grade components, to boot. It's unclear at the moment if I would have as easy a time using a desktop from the time as completely as I can with this system, but it's clear that given an opportunity, I can easily find a use for as much computing power as I can have handy. There's no reason to believe I couldn't do the same (or more) on a bigger and better computer, however.
And, in the end, there's not even any aesthetic reason for anybody else why it would be a bad idea for me to own a big and over-powered computer. It would be hidden away in my own office or bedroom.