Wang Jin-dong1, Rong Xiao-xia2
1 Shandong Computer Science Center, Jinan,
2 School of Mathematics and System Sciences, Shandong Univversity, Jinan,
A data stream management system accepts push-based inputs from a set of data source outside and processes these inputs with respect to a continuous query network. When input rate exceeds system capacity, the system will overload and latency will deteriorate.
In the system with multi-input and multi-output data streams and terminating with an aggregate operator, how to solve load-shedding problem is never studied before. Aiming at this new problem, a load-shedding algorithm based on sub-path is proposed. According to this algorithm, continuous query network is divided into separate sub-paths. Based on which, load is calculated as the sum of each sub-path load and samplers are set one sub-path after another. At last, a simulation was made to compare this algorithm with simple load-shedding algorithm by discussion the relationship of query error and load value, time and load bursting. The algorithm can improve the query precision in some degree.
Key words: Data stream, Continuous query, Load shedding, Sub-path