An Improved Long-Term File-Usage Prediction Algorithm

Abstract

As more computing centers collect files to use in data-mining or datamarts, managing long-term storage space becomes more important. We describe a new way to more accurately predict which files will be used in the future. This new method is an order-of-magnitude more accurate than any current technique. Fifty to eighty percent of all user files can be compressed or moved to tertiary storage with little impact on the user perceived performance. This paper supports these conclusions with an analysis of long-term (5-8 months) data collected on different types of computing environments.