Every now and then, we have 'interesting' conversations at work about morbid subjects; recently we had a discussion about another 9/11 and whether our systems could cope with it. Not the systems that you get to see but our internal systems and how we scale for peak load and what peak load actually means. Scaling web-servers is pretty easy but other systems don't necessarily scale so easily.
We had recently put a new system in and then Michael Jackson went and died; so that was a good test for the system, which coped pretty well. Tragic events are what generates traffic; that and England winning the World Cup but we have generally have to deal with tragedies not miracles.
But web-sites, phone-systems, networks can all fail when under extreme load; in fact, pretty much every form of utility is massively over-subscribed pretty much by design. If everyone got in their cars at the same time, the road network would melt-down. Flights are regularly over-sold, certainly by the budget providers. If everyone in the UK decides to make a cup of tea at the same time, the power-grid suffers.
This led me on to thinking about the Public Cloud; what level of over-subscription do the Public Cloud providers sustain in the forms of 'reserved instances' and how many instances can they actually support all coming up at one time. How much storage does Amazon actually have?
If you are going to bet your business on the Cloud, you probably ought to know and you probably ought to know what kind of events will cause you to burst and probably what will cause everyone else sharing the Cloud with you to burst. And what you are going to do if the Cloud does burst? Do you know where your umbrella is? Or at least the candles to dry you out.
