Recently a new class of data streams applications has become widely recognized: applications in which the data is multiple, continuous, quickly changed with time. The access to data stream is also continuous and many-times, and requires on-time response. Examples of such applications include network traffic monitoring, telecommunications data management, sensor networks, and others. The experience shows that the traditional RDBMS cannot satisfy the data stream system's requirements. So how to design an effective new database system in order to process data streams becomes an open research problem. The purpose of this dissertation is to design a universal data stream management system not only for data streams but also for static relations. Today there is a little work about this in our country, we designed a data stream management system, named Phosphor,it refers to a few classic data stream management systems. The important difference between Phosphor and other DSMSs include: both data streams and static relations are the management objects of Phosphor, continuous query language with comprehensive window types, the novel technique for organizing data streams in memory, the strategy of load shedding based on memory resource.
|