Control-Based Load Shedding in Data Stream Management Systems

Load shedding has been widely used in data stream management systems (DSMSs) as a means to maintain Qualityof- Service (QoS) of data processing. However, current systems tend to adapt ad hoc strategies in determining the timing and magnitude of load shedding. In this paper, we argue that these strategies cannot effectively maintain QoS due to the existence of environmental/internal dynamics. To remedy this problem, we propose a systematic approach that takes advantages of well-established feedback control techniques. Our approach can achievemuch less QoS violations with the same amount of data loss. Initial experiments performed on a real-world DSMS support our conclusion.