Continuous Query Processing With Deadline Over Streams

With the emergence of data streaming applications, such as the financial analysis, network monitoring, and sensor data monitoring and controlling, a new data management technology, query processing over data stream, has obtained a very large development. Currently, the research of data stream technique focuses on the general real-time applications. In some applications that have a higher real-time requirement and have a certain critical degree requirement, such as the real-time spacecraft control systems, on-line auction system, etc., query processing should be completed within a certain time limit. If the deadline is missed, it may result in very serious consequences. The deadline and the high critical task for such applications are two prominent factors. The data stream processing technique based on the combination of these two factors with some other relevant factors is needed to improve query processing efficiency of high real-time and high critical applications.In this paper, a priority calculation method that involves comprehensive multifaceted factors including the deadline, idle time and critical degree is proposed, making the earlier deadline, shorter idle time and higher critical degree, the higher the query priority. The linked list and the sort ideas are used to solve priority value's uniqueness problem. Taking into account the basic characteristics of the data flow velocity, based on the "time window" concept, the batch processing ideas are used to handle data stream tuples, to reduce the scheduling switch overhead and to improve system performance. The proposed method is implemented in the STREAM prototype system. The experimental result shows the effectiveness and usability of the method.