Here's an example of the kind of comment we are receiving.

Berry van der Linden:

We actually check for latency before the update if it doesn't meet out threshold it will cause an error, if this happens 5 times an engineer gets a warning message. We also check during update for latency issues. In our experience latency is usually resolve within our 5 checks.

Cost can be high but we find it is necessary to have back-ups done this way. We charge the client for the cost and they agree that the loss of data would be worse.

The key for us is limited downtime, because even a good cloud can go down because of a hardware problem.

Keep the comments coming. You'll need a Disqus account to participate.