The First Principle of Big Memory Systems

TL;DR

This paper emphasizes persistence as the core principle of big memory systems, analyzing their hierarchy extensions, and proposing design methodologies for efficient, high-performance, and cost-effective persistent memory architectures.

Contribution

It introduces a comprehensive analysis of big memory system hierarchies and proposes novel design methodologies including speculative and deterministic persistence.

Findings

Networks flatten storage hierarchies

Persistence is fundamental to big memory systems

Proposed methods improve cost efficiency and performance

Abstract

Persistence is the first principle of big memory systems. We comprehensively analyze the vertical and horizontal extensions of existing memory hierarchy. Networks are flattening traditional storage hierarchies. We present the state-of-the-art studies upon the big memory systems, together with design methodology and implementations. We discuss the full-stack and moving persistence. In order to achieve cost efficiency and deliver high performance, we present the speculative and deterministic persistence.