Research And Design Of Data Stream Management System

Recently there are new data-intensive applications, such as financial service, telecommunication data management, network monitoring, sensor detection, and so on. Such applications have similar features: data is instantaneous, continuous, and time-varying. The data is more suitable for being modeled by transient data streams than by permanent relations. Putting such continuous arrival data into the traditional relation database to manage is not feasible. Then there are some new researches on data stream.In this paper the emphasis is the research and design of the data stream management system. We analyze the foreign major existing data stream management systems, compare the difference between traditional database management system and the data stream management system, design the framework of the data stream management system MyStream, give the detailed descriptions of each module and the data structure descriptions of the key parts, propose a new SQL-based data stream query language MyCQL, use the COM interface in the user interface, propose an improved data stream timestamp management algorithm, adding probability to the original STREAM timestamp management algorithm. The algorithm makes the parameter settings more flexible. In the operator schedule aspects, we propose a rate-based preemptive batching scheduling algorithm, give some definitions of the rate computation, and compare with other scheduling algorithms such as FIFO, RR. We apply MyStream to the traffic control, set the different fees according to the traffic conditions to reduce the traffic congestion.