Print

Tran, Tri Minh

Computer Science

2010

PhD

A data stream is a continuous unbounded sequence of data items. In many stream applications (e.g., network monitoring, sensor network monitoring, online auction monitoring), join queries (Le., queries involving join operators) are one of the most common types of queries. Due to the unbounded and continuous nature of data streams, however, processing this type of queries over data streams is very different from processing them over databases, and this difference brings numerous challenges. This journal-format 'dissertation addresses three problems important to improve the efficiency of stream-join query processing: (1) efficiently processing join queries when aggregation operators are in the query as well, with the focus on a new query transformation rule to derive more efficient query execution plans, (2) efficiently processing join queries in a distributed environment, with the focus on utilizing semi-joins to reduce the communication cost, and (3) efficiently processing join queries by continuously adapting the query execution plan to the changes in a distributed environment, with the focus on detecting the changes locally at individual nodes and generating and gracefully migrating to more efficient execution plans.