Uptime is a bad proxy for availability

Is it equally important for your service to be functioning optimally at 3 AM on a weekend as on a Tuesday afternoon, after it’s just been mentioned in a press release and product usage is soaring?

Is it equally important to you that you can set the preferred language or timezone in your banking app as that you can actually pay at the check-out in a store?

More than likely, you’ll agree with me that these “service levels” aren’t of equal importance, which is why it doesn’t make a lot of sense to use the concept of “uptime” as the core metric to determine system availability.

A better alternative is to express system availability as a ratio of events.