BACKGROUND Hierarchically arranged memory has been a common feature in computing
for some time. Fundamentally, faster memory is more expensive per byte. Despite rapid
advances in storage performance, it is often economically unsound to utilize only the lowest
latency storage medium. Instead, in order to deliver acceptable performance within a fixed
budget, storage devices of different sizes and speeds may be arranged so that memory
transactions read or write to the fastest devices whenever possible. In a typical example, a …