Control-based quality adaptation in data stream management systems

Abstract

Unlike processing snapshot queries in a traditional DBMS, the processing of continuous queries in a data stream management system (DSMS) needs to satisfy quality requirements such as processing delay. When the system is overloaded, quality degrades significantly thus load shedding becomes necessary. Maintaining the quality of queries is a difficult problem because both the processing cost and data arrival rate are highly unpredictable. We propose a quality adaptation framework that adjusts the application behavior based on the current system status. We leverage techniques from the area of control theory in designing the quality adaptation framework. Our simulation results demonstrate the effectiveness of the control-based quality adaptation strategy. Comparing to solutions proposed in previous works, our approach achieves significantly better quality with less waste of resources. © Springer-Verlag Berlin Heidelberg 2005.