The problem of dynamically adjusting the sizes of sliding windows when the rates of data streams or continuous queries change in data stream systems is studied in this paper. Based on the amount of available memory resource and the requirement of queries, three classes of algorithms for dynamically adjusting the sizes of sliding windows are proposed. These algorithms provide three levels of quality of service to all kinds of continuous queries and enhance the efficiency and effectiveness of processing continuous queries. Analytical and experimental results show that the algorithms can be applied to data stream systems effectively.