Locality in Computer Archtecture
Peter J. Denning. The Locality Principle. In Communication Networks and Computer Systems (J. Barria, Ed.). Imperial College Press (2006), 43-67.
CHAPTER 4
The Locality Principle
Peter J. Denning
Naval Postgraduate School Monterey, CA 93943 USA Email: pjd@nps.edu Locality is among the oldest systems principles in computer science. It was discovered in 1967 during efforts to make early virtual memory systems work well. It is a package of three ideas: (1) computational processes pass through a sequence of locality sets and reference only within them, (2) the locality sets can be inferred by applying a distance function to a program’s address trace observed during a backward window, and (3) memory management is optimal when it guarantees each program that its locality sets will be present in high-speed memory. Working set memory management was the first exploitation of this principle; it prevented thrashing while maintaining near optimal system throughput, and eventually it enabled virtual memory systems to be reliable, dependable, and transparent. Many researchers and system designers rallied around the effort to understand locality and achieve this outcome. The principle expanded well beyond virtual memory systems. Today it addresses computations that adapt to the neighborhoods in which users are situated, ways to infer those neighborhoods by observing user actions, and optimizing performance for users by being aware of their neighborhoods. It has influenced the design of caches of all sorts, Internet edge servers, spam blocking, search engines, e-commerce systems, email systems, forensics, and context-aware software. It remains a rich source of inspirations for contemporary research in architecture, caching, Bayesian inference, forensics, web-based business processes, context-aware software, and network science.
1. Introduction Locality of reference is one of the cornerstones of computer science. It was born from efforts to make virtual memory systems
CHAPTER 4
The Locality Principle
Peter J. Denning
Naval Postgraduate School Monterey, CA 93943 USA Email: pjd@nps.edu Locality is among the oldest systems principles in computer science. It was discovered in 1967 during efforts to make early virtual memory systems work well. It is a package of three ideas: (1) computational processes pass through a sequence of locality sets and reference only within them, (2) the locality sets can be inferred by applying a distance function to a program’s address trace observed during a backward window, and (3) memory management is optimal when it guarantees each program that its locality sets will be present in high-speed memory. Working set memory management was the first exploitation of this principle; it prevented thrashing while maintaining near optimal system throughput, and eventually it enabled virtual memory systems to be reliable, dependable, and transparent. Many researchers and system designers rallied around the effort to understand locality and achieve this outcome. The principle expanded well beyond virtual memory systems. Today it addresses computations that adapt to the neighborhoods in which users are situated, ways to infer those neighborhoods by observing user actions, and optimizing performance for users by being aware of their neighborhoods. It has influenced the design of caches of all sorts, Internet edge servers, spam blocking, search engines, e-commerce systems, email systems, forensics, and context-aware software. It remains a rich source of inspirations for contemporary research in architecture, caching, Bayesian inference, forensics, web-based business processes, context-aware software, and network science.
1. Introduction Locality of reference is one of the cornerstones of computer science. It was born from efforts to make virtual memory systems