Abstract: The current dynamic materialized view selection algorithms based on disk cache can tune with the altering query requirements and possess better dynamic adaptability.However these algorithms simply refer to the memory mechanism, whithout an accurate model to take into account the cost of retrieving materialized views from disk.A dynamic materialized view replacement (DMVR) algorithm is presented, which is based on a better cost model and therefore suitable for dynamically adjusting the materialized view set.For systems with a heavy workload, that DMVR algorithm is modified to become DMVR-IDLE, which possesses and maintains a better response performance to high frequency query.Experiments have shown effectiveness of both algorithms.

Key words: cache mechanism, data warehouse, materialized view