Fido: A Cache That Learns to Fetch

This paper describesFido, a predictive cache [Palmer 19901that prefetchesby employing an associativememoryto recognizeaccesspatterns within a context over time. Repeatedtraining adaptsthe associativememory contentsto data and accesspattern changes, allowing on-line accesspredictions for prefetching. We discuss two salient elementsof Fido - MLP, areplacement policy for managing prefetched objects, and Estimating Prophet, the component that recognizes patterns and predicts access. We then present some early simulation results which suggest that predictive caching works well andconcludethat it is apromising method.

• Type: Conference
• Year: 1991
• Where: VLDB
• Authors: Mark Palmer, Stanley B. Zdonik
• Keywords: Associativememory Contentsto Data, Database, Predictive Cache, Salient Elementsof Fido, VLDB 1991