This session mixes many of the lesson learned in years while writing distributed software. First is "design for failures", to be sure our code will never generates inconsistencies. Second is "stateless (almost) everywhere", to be sure our compute power can be built in minutes. Third is "performance", which has to be considered a feature itself while it is often a competitive gap between you and others. In this session we approach the "simple" problem of reducing the computation time of a CPU intensive application, using C#, Queues, Azure Container Instances and a bit of ARM templating.