Incremental Continuous Query Processing over Streams and Relations with Isolation Guarantees

摘要

Stream processing has become an important research issue with the increase in stream data sources. Many stream processing systems need to reference non-streaming resources such as database relations to answer real world queries. Since database relation is a shared entity, it may be updated during the continuous query (CQ) execution by other database clients resulting in inconsistent query results (partly using the relation before update and partly after update). For this problem, an isolation model is needed to define the way in which these updates are reflected in the output of the stream-relation join. In this work we propose an incremental CQ processing approach with isolation guarantees which makes use of a monitor operator to transform the relational updates into stream tuples. Since database relations tend to be large, an in-memory T*-Tree index is used to increase the stream-relation join efficiency. Experiments are performed to prove that guaranteeing isolation solves the inconsistency problem and to show that the incremental computation and indexing improves the query throughput significantly.