Don't panic about cloud based application failures

“Gmail isn’t working – OMG.  Call out the National Guard”

The recent discussion around the failure of some notable cloud based web services has reminded me of a comparable discussion in the transportation industry.  The number of fatalities associated with automobiles greatly exceeds those associated with air travel.  Why is that people assume otherwise?  Because when a plane crashes to the ground the loss of life is sudden and dramatic.  When an individual dies in an auto accident we are conditioned to view it as less significant.

What’s the point?  Gmail failed last week.  Amazon S3 failed the week before.  Twitter fails weekly (or so it seems).  These are large public outages which drive a lot of media coverage.  On the other hand, the daily (or more likely hourly) outages individuals encounter with their own tool set receive little or no media coverage. There is no way to measure them.  There is no way to report on the collective impact on productivitity associated with thousands of individual mail servers being crushed under the weight of spam.

Large scale service delivery firms that focus on service excellence will eventually drive down these failures.  You and I will continue to build things that break.  Oh well.