Predictive Caching Using the TDAG Algorithm

Abstract We describe how the TDAG algorithm for learning to predict symbol sequences can be used to design a predictive cache store. A model of a two-level mass storage system is developed and used to calcdate the performance of the cache under various conditions. Experimental simulations provide good confirmation of the model. Introduction A cache is a multi-tier store consisting of a hierarchy of storage media with different access speeds, capacities, and costs. The medium with the highest capacity is simultaneously the one with the least cost per byte and the slowest data access rate. In a typical mass storage system (MSS), for example, the majority of archival data is kept on tape, while the portions in active use are retained on disk or in RAM. A request for a segment of data that is not available in the higher-speed store (or cache) is answered by replacing one of the current segments in the cache by the requested segment. Moreover, if the segment being replaced has been modified,