Abstract

This paper examines the costs and potential benefits of long-term prefetching for content distribution. In contrast with traditional short-term prefetching, in which caches use recent access history to predict and prefetch objects likely to be referenced in the near future, long-term prefetching uses long-term steadystate object access rates and update frequencies to identify objects to replicate to content distribution locations. Compared to demand caching, long-term prefetching increases network bandwidth and disk space costs but may benefit a system by improving hit rates. Using analytic models and trace-based simulations, we examine several algorithms for selecting objects for long-term prefetching. We find that although the web’s Zipf-like object popularities makes it challenging to prefetch enough objects to significantly improve hit rates, systems can achieve significant benefits at modest costs by focusing their attention on long-lived objects. 1

Citations