Research On Query Optimization Technology In Distributed Real Time Database

Real-time database system(RTDBS) is the integration of real-time system and database technology. Currently, the real-time database system is widely applied to military,telecommunication,electricity,aviation,industry control and some other domains, the most of these domains are distributed, therefore, the research of distributed real-time database system catches the attention of experts in real-time field,database field and distributed field. DRTDB becomes a new research hot point.China has a large area of ocean, the marine environment is very important for the development of her economy. Marine environment online monitoring is proposed for the requirement of marine disaster alarming, in which the necessary data information of marine is provided by real-time monitoring system in ocean environment. This thesis takes "the marine environment online monitoring and marine disaster intelligent forecasting system" as a background, researches the construction of the common frame of the distributed real-time database, points out the DRTDBS' functional characteristic and key technology, and proposes a distributed real-time database construction scheme for the actual circs of the marine environment, which will satisfy the requirement of whole system's real-time data management and information promulgation.The data management subsystem of marine monitoring system has the character of distributed and real-time, local real-time databases transmit real-time data and command through wireless network. For the restriction of transmit rate of wireless network, it is required to develop an effective query processing method to reduce the query response time, so as to match the system's real-time ability. After analyzing traditional query optimization strategy, this thesis points out the merit and the drawback of the traditional strategies and proposes a new distributed query method, the up-down semi-join strategy based on three query. The new query method proposes the concept of strong semi-join, in which multi relations' restriction were used to improve the reduction of one relation, the method also proposed an up-down reduce strategy, which can lead the relations in a tree to a full reduction state, therefore, the useless records are eliminated before transmission and the query's response time is reduced. The thesis also researches the full reduction strategy when there are two or more join attributes between two relations and the full reduction strategy of cyclic query, proposes the conception of default join restriction, points out the inherent difference between tree query and cyclic query, and propose a new two-way semi-join method based on additional attribute to convert a cyclic query to tree query.Relative to the FRD method, up-down reduction strategy can also reach the full reduction state, but can further reduce the data transmission cost of the down phase, especially to the relation that has many connect attributes, thereby reduce the query response time.Real-time query must match its time requirement, or disaster may occur, the thesis researches the fault-tolerance scheduling algorithm of the periodicity real-time transactions, and proposes a preemptive estimate algorithm, which can reduce the preemption to the low priority task that will finish soon so as to decrease the waste of the CPU resources, the new algorithm also use a load balance strategy to optimize the task scheduling more. By the experiment, the new algorithm can decrease the CPU resources waste and improved the success rate of system's real-time task.