Research And Realization Of Data Stream Management System And Its Correlated Algorithms

Traditional database management systems are designed to handle persistent data andanswer transient queries. However due to the evolution of network, telecommunication andsensor technologies, a new data processing model has recently arisen. In this scenario,continuous and long-running queries are posed over transient streaming data. TraditionalDBMS are not designed for rapid and continuously loading of individual data items, and theydo not directly support the continuous queries. Limitations of traditional DBMS in supportingstreaming applications have been recognized, prompting research to augment existingtechnologies and build new systems to manage streaming data.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 accessto data stream is also continuous and many-times, and requires on-time response. Examples ofsuch applications include network traffic monitoring, telecommunications data management,sensor networks, and others. The experience shows that the traditional RDBMS cannot satisfythe data stream system's requirements. So how to design an effective new database system inorder to process data streams becomes an open research problem.There are some improvements in its correlated algorithms of the data streammanagement system aimed at the unstable network rate in this paper in order to have a rapidand timely response to the treatment as speed makes a sudden increase. This dissertation alsoscheme out a universal data stream management system, named PowerStream, not only fordata streams but also for static relations. Today there is a little work about this in our country,and it refers to a few classic data stream management systems. The important differencebetween PowerStream and other DSMSs include: a better adaptive capacity to unsteadiness ofthe real speed of the network, both data streams and static relations are the managementobjects of Phosphor, continuous query language with comprehensive window types, the noveltechnique for organizing data streams in memory, the strategy of load shedding based onmemory resource.