001 Hosting

"Our own server is cheaper" is true until the disk dies

The bill for one machine is small. The bill for the week it stops is not. What renting hosting actually buys, and the question underneath the choice.

“Our own server is cheaper” is usually true. It is also the whole of the argument, which is the problem.

The bill for one machine is small and arrives every month, so everyone has seen it. The bill for the week it stops is large, arrives once, and nobody has seen it yet. Setting one against the other is not a cost comparison.

What renting actually buys

Nothing magical. Three dull things:

  • A machine you can replace. Not repair. Replace, in minutes, without anybody driving anywhere or sourcing a part.
  • A copy of the data that is not on that machine. Worth stating plainly, because a copy on the same disk is not a copy.
  • Somebody whose job is installing the security fixes. Not somebody who does it when there is time.

You can build all three yourself. People do, and some of them do it well. This is not an argument that one machine is wrong.

The question underneath the decision

It is not a technical question. It is whether the person who knows how to rebuild that machine should be the same person you call when a customer is on the phone.

On one machine, those are the same person. The day you need both at once is the day you find out what that costs. That is what somebody means by blast radius: how much of the business stops when one thing stops.

The failure is quieter than you expect

The version that goes wrong does not announce itself.

A backup routine writes into a folder that was deleted years ago and reports success every night. The job is green. The log says done. The only way to find out otherwise is to ask for a file back, and nobody ever has. That is a real one we wrote up — 640 gigabytes of customer files on a single machine, with a nightly job that had been succeeding into nothing.

The same quiet applies to updates. We moved a system off a machine that could not be updated: three years with no security fixes, a disk 96% full, and an office where nobody knew any of it until somebody looked. Nothing had broken. That is the normal state right up until it isn’t.

If you are weighing this now

Two questions settle most of it. When did somebody last restore a file from your backup and confirm it opened? And if the machine were gone this afternoon, who would rebuild it, and what would they need that is not on it?

Neither needs a consultant to answer. If both answers come back clean, keep your server — it probably is cheaper.

If you are about to make this call and want somebody to argue the other side, that is a conversation we are happy to have, and what things cost is published so you can see the shape of it first.

Recognise any of this in your own system? Describe what breaks and a person replies within one business day. Free, and a person is what’s at the other end.

Write to us  ·  The work  ·  What things cost  ·  All writing