Including DB & API machines, our web app occupies 18 heavily utilized Amazon EC2 instances under a traffic load of ~7k RPM. "Conventional wisdom" says that we're mis-using EC2 by keeping so many virtual instances running but I prefer hard data to "conventional wisdom".<p>Curious to hear from the HN community: what’s the most long-running EC2 instances you've seen in a successful web app?
Some other data points:<p>* Animoto was running several thousands of machines in 2008 (<a href="http://bit.ly/EDLtt" rel="nofollow">http://bit.ly/EDLtt</a>)<p>* Litmus runs 400 servers (<a href="http://bit.ly/d7Hc7y" rel="nofollow">http://bit.ly/d7Hc7y</a>)<p>* 99Designs runs entirely on EC2 (<a href="http://bit.ly/aotKgg" rel="nofollow">http://bit.ly/aotKgg</a>)<p>I've personally had long-running EC2 instances with uptimes in <i>years</i>. You could do it cheaper in terms of hardware, but at the cost of wasting time at the colo while you could be building cool shit.
Zynga runs over 12,000 EC2 nodes total. A single app like FarmVille, which runs entirely on EC2, will be a few thousand. It's safe to guess we're Amazon's biggest customer.