How would you design a highly available, fault-tolerant infrastructure for a global service at New Relic?
