How would you design a system that is scalable, fault-tolerant, and highly available at Amazon Services?
