Choose Between Two Machines and Allocate Work from Performance Curves

A capacity planner must choose one hardware type to serve a steady request stream. Two types are on offer, A and B, and each is described by three numbers: fixedMs, the milliseconds a machine burns before it starts on a batch; perReqMs, the extra milliseconds every request in the batch adds; and pricePerHour, the rental cost of a single machine. Requests are handled in full batches only, and the permitted batch sizes are given in batchSizes. The stream arrives at lambda…