Thinking about Availability in Large Service Infrastructures

   Abstract