Application Research On Load Management Technologies Of DSMS

With the development of computer, network, and the relevant technologies, information systems are widely used in manufacturing, transporting, finance, IT and other regions. It makes traditional data management methods not applicable to the demand of new time for the rapid increasing of data amount. In recent years, researchers begin pay more attention to data stream management technologies, such as constructing and optimization of data stream management system (DSMS in short), data stream mining and so on.In this thesis, we study the basic structure, principle, realization, characteristic and the main application fields. To deal with overload problems aroused by the variety of input data rate, we discuss the method of constructing load management system, further more study particularly the key technology of load-balancing, load-shedding and data summary.Main contributions of this thesis are the following:(1) A load-balancing algorithm is proposed based on two-tier structure. To satisfy the demand of expansibility and refresh ability about distributed data stream management system, we studied the load-balancing technology and propose the algorithm, which can balance load of tightly-coupled distributed DSMS in Internet(2) A load-shedding algorithm is proposed based on sub-path analysis. To solve the problem of load shedding about continuous query network in which has multiple input and output data streams and terminated with an aggregate operator, we studied the load-shedding technology and propose the algorithm. On this bases, two improved algorithms are given in cases such as input data items with different weight, continuous query networks terminated with non-aggregate operator. They extend the applicability of the algorithm.(3) An efficient discovery method for useful data items is proposed. At first the definition of useful data items in distributed DSMS is given, then the discovery method of single input data streams is extended to multiple environment and get the algorithm, which can solve the problem efficiently.