Load Shedding in XML Streams

摘要: Because of the high volume and unpredictability arrival data streams, stream processing systems may not always be able to keep up with input — resulting in buffer overflow uncontrolled loss data. Load shedding, prevalent strategy for solving this problem, has todate been considered relational engines. On other hand face additional challenges opportunities ”structural shedding”, due complex nested XML result structures. We now tackle open shedding problem by a three-pronged solution. First, we develop preference model XQuery enable users specify relative importance preserving different subpattern structure. This transforms into rewriting user query possibly several queries that return approximate answers yet highest possible utility as measured given model. Two, cost compare both performance alternate queries. Third,we propose two solutions: OptShed, FastShed. OptShed guarantees find an optimal solution however at exponential complexity. FashShed confirmed our experiments, efficiently achieves close-to-optimal wide range cases. Lastly describe in-automaton mechanism Raindrop system. The experimental results show proposed preference-driven solutions consistently achieve higher compared existing “relational” techniques.