No, it's not Dell versus HP. Or Apple versus everyone else. It's cheap, commodity hardware versus expensive, specialized hardware.

This has always been a point of debate in any sort of hardware procurement decision, and you've probably guessed by now that I come down on the cheap, commodity side of the lines eight times out of ten.

Tom's Hardware explores the evolution of the question as it is currently being applied in the Web 2.0 era, with massive datacenters becoming the focal point of IT services, and the servers that populate them being the hardware in question. In particular, they look at the specialized solutions being offered by the likes of HP and IBM and contrast them to Google's "lots of generic servers in lots of rooms" approach.

There is a lot of good technical detail in the article about how the specialized approach can address some of the common problems found with commodity hardware, but that strikes me as being beside the point. It's not as if any of this stuff started off as a commodity, after all. Big, specialized servers were the rule for many years, and in fact were all that were available for much of that time. Moving to commodity boxes is the alternative, not vice versa. And it was done, and still is, because despite the advantages the generic offerings are still more cost-efficient than the specialized hardware in many cases. IBM and HP have clearly done their homework, and I have no doubt what they are selling works as advertised to improve efficiency, but it does so at additional cost and complexity which may continue to outweigh the benefits in many scenarios.