A fairly significant outage for Amazon EC2 in North American over the past week has generated a pile of press coverage – mostly, it has to be said, because the incident took out a number of high-profile social networking sites including Quora and FourSquare.
There are a few things I take away from this:
1) social networking sites go down – world keeps turning
2) single points of failure are a bad thing
3) this incident will be used by all and sundry as a reason why Cloud is bad, and they are wrong
I am reminded of a comparison that I heard (I think) Geoffrey Moore make at an event back in 2009. He saw commonality in Cloud and the airline industry in what he unfortunately called their propensity to “kill people in batches”. Air travel is remarkably safe; however, when it goes wrong, it does so in a way that has a significantly newsworthy impact (excuse the pun).
So, it seems, it is for Cloud service providers. The thing is that we humans are actually very poor, for the most part, in assessing risk and probability. If we were generally any good at it then the insurance industry wouldn't exist (nor the gambling industry for that matter). Couple that with something new, and all sorts of daft conclusions will be drawn from over-publicised incidents. (On the Risk thing, take a look at Dan Gardner's excellent book on the topic.)
I was in conversation with the lady who is responsible for the Microsoft UK “Cloud Power” marketing campaign last week. I guess that she is a fair few years younger than me, so listened in disbelief as I told her that in my first job (from 1993-1995) I had an email account, but no Internet email address for people outside of my then employers to contact me. It just wasn't done – if you wanted to get an externally-accessible account then you needed forms filled in in triplicate from the appropriate managers. The risks of external email were too great to be given to all and sundry.
This sounds somewhat outlandish these days. In the same way, probably, that more and more services moving into the Cloud seems to some in the IT industry at the moment, particularly in the aftermath of the Amazon failure.
But let's come back to the airline analogy. Putting aside that many of the services running on Amazon (nb I'm duty bound at this point to say that other Cloud platforms are available) wouldn't exist if it weren't for Cloud hosting, let's imagine that all of the things running on Amazon had been instead on home-grown, in-house infrastructure. What would have been the collective down time for all of the individual services' own hosting up until the point of failure last week? Hypothetical, but if it helps by way of comparison in 2004 under 800 people were killed in airline incidents; the same year saw 1.2 million people die on the roads across the globe. Yet, how many people find flying scary in comparison to getting in a car?
mmitII 