Abstract:
Superpages (2MB pages) can reduce the address translation overhead for large-memory workloads in modern computer systems. This paper clearly outlines the sequence of events in the life of a superpage and explores the design space of when and how to trigger and respond to those events. This provides a framework that enables better understanding of superpage management and the trade-offs involved in different design decisions. Under this framework, this paper discusses why state-of-the-art designs exhibit different performance characteristics in terms of runtime, latency and memory consumption. This paper illuminates the root causes of latency spikes and memory bloat and introduces Quicksilver, a novel superpage management design that addresses these issues while maintaining address translation performance.
