If you’ve got access to all that shit, why not host yourself.

The last large installation I was a sysadmin for had somewhere in the neighborhood of 500k sq/feet of raised floor. Upon that sat thousands of machines of various shapes and sizes.

Nothing really failed all that often. When we had an actual hardware failure that needed something swapped out (not talking about drives in RAIDs - those go all the time and are hot-swapped out), it was a rare occurrence. Non-redundant components (memory, network controllers, motherboards, etc) didn’t simply fail on a regular basis like they do here (as this blog shows).

Now … yes, that was HP, Sun, and SGI enterprise hardware and not off-the-shelf x86 stuff … but still. As another comparison, I now work at a start-up doing server-side development. We have maybe 30 x86 based servers (mostly Sun, some Dell), and a couple big Sun boxes for the databases … and yet, they don’t seem to crash or fail either. And they’re pretty hammered 24/7.

So yes, I see their fail rates as high. I would suspect it’s because they use cheap hardware ($10 a month doesn’t buy you ultraSparcs) and their infrastructure does seem to suck when their whole network goes to hell on a monthly basis. Other than the fact that it breaks so much, I don’t have a real insight on how they do things … I know I personally wouldn’t do and rely on as much NFS as they do for everything, but that’s just me - I’ve made that mistake before.

Well, “alternate realities” and “parallel” universes aside, I’m willing to bet we both live in the *same* world - though it is not surprising that we may view it differently. I’m curious as to what metrics you are able to apply to how things break” at DreamHost. Unless you are privy to a lot more detail about the infrastructure at DreamHost than they have ever made public, it seems to me that you have no way to determine what their “hardware failure rates” actually are, and are therefore unable to objectively compare them at all to “the rates that (you’ve) experienced as a professional sysadmin and software engineer over the last 15 years,” let alone pontificate that their ” rates are somewhere in the neighborhood of 10x - 20x the rates that (you’ve) experienced.”

I agree that, while costing more, the “DreamHost of old” was more reliable, and some stuff that I would have hosted in those days is no longer suited for today’s DreamHost. That said, I don’t consider myself to be “over-zealous” when I maintain that many *are* well served by DreamHost, and report that *I* am one such customer. That fact that YMMV is normal, as is that fact that I may perceive *your* views to by hyper-critical.

Hopefully, some host will embrace a model that is profitable, and allows growth, that is more like the days of the DreamHost of old for your (and some of my) sites that are better suited for a more robust, yet less expensive that a dedicated server, hosting product. There are certainly some of my sites that would benefit from that.

None of that, however, changes the fact that for many of my clients (yes, even business clients) DreamHost remains a phenomenal value (”warts”, and “broken stuff,” and all).