An analytical approach to file prefetching

flei djdgcscolumbiaedu File prefetching is an eective technique for improving le access performance In this paper we present a le prefetching mechanism that is based on online analytic modeling of interesting system events and is transpar ent to higher levels The mechanism incorporated into a clients le cache manager seeks to build semantic structures called access trees that capture the corre lations between le accesses It then heuristically uses these structures to represent distinct le usage patterns and exploits them to prefetch les from a le server We show results of a simulation study and of a working im plementation Measurements suggest that our method can predict future le accesses with an accuracy aroun