One of my tasks over the weekend was to calculate how much it would cost to operate a service. I can't tell you what the service is, of course, since actually providing useful context is far beyond the scope of this blog. ;-) However, I had a good sit-down with Excel 12 on Saturday morning and began refactoring the way I think into things like "hardware" and "unanticipated downtime." I was struck by a few things:

  • Excel 12's table autoformat feature doesn't work the way I do; this caused me no end of grief and my tables still look weird.
  • People say "hardware is cheap," and maybe it is, but it sure adds up. Sure other costs eventually outweigh hardware, but good quality rack-mounted web servers are still expensive, especially when you're buying double to allow for full redundancy.
  • You don't have to allow for full redundancy. I don't know where the breakeven point is exactly, but 100% redundancy isn't really necessary.
  • The service time to replace a hard drive will cost more than the hard drive itself.
  • Most people don't know how much it costs to operate their sites. With the exception of some of the big sites, which know to-the-penny how much things costs, most of the site operators I've spoken with (even for sites with millions of unique users) only know a rough annual approximation of hosting costs, and even then don't account for everything.

I'd love to say that I now have a spreadsheet that captures every variable and gives phenomenally accurate predictions of cost. Instead I have a spreadsheet that takes about 6 variables and calculates an annual operating cost for a particular kind of site. Interesting, but not the panacea I'd hoped for.