Design And Realization Of Online Cluster In Workload Adaptation Database

With the development of Internet and information technology,the database management system becomes more and more complicated,it is huger and huger that its data store the scale,and become more and more expensive and tedious to manage the database system too. Under the traditional way, it is mainly the craft regulation of depending on the system manager. But it is unable to carry on the dynamic regulation to the situation of utilization of resources to do it in this way;At the same time the database administrator's expenses bring the heavy burden for the maintenance of the database system too. The autonomictic database system can manage oneself automatically and voluntarily, reduce the complexity of artificial data base administration,thus the extra cost brought while reducing the manpower cost and system trouble. So,the independent technology of the database,for improving the application ability in the key is used of database system,reduce overall possession cost , database of system ,reduce hardware, investment of softwarehave important meaning.Professor Niu BaoNing of the Taiyuan university of technology have already put forward the framework for Workload. This framework is conposed of the Workload control progress and workload detection progress,the processes are in turn made up of four functional components -workload characterization, performance modeling,workload control, and system monitoring. In this framework, the system can predicte and control workload according to the performance of some workload.In the workload adaptive database system, workload characterization is the key part, firstly, it should classificate,then predict workload performance according to the situation classified.this text is mainly research classification. Because the classification of workload requires the real-time online, it must be simple and swift. So, this text proposes a cluster based on prioritized incremental algorithm.Firstly,this text has introduced the research background of this subject, namely:The development of the adaptive database and state of studying, and the data mining some basic situations of technology.sencodly, this text has introduced cluster's technology, including the definition of the cluster, the type of the data, some cluster algorithms .Thirdly,the text introduced K-means algorithm and the incremental cluster algorithm,nd their thought,mputational process,dvantage and disadvantage.Finally, this text have introduced on the cluster based on feature vectors and increasemental cluster based on prioritized ,and has verified the validity of above-mentioned two algorithms with the experiment.The innovation of this text has using the increasemental cluster based on prioritized,he prioritized in this algorithm refers to the result cluster based on feature vectors .That is to say , the increasemental cluster based on prioritized based on cluster based on feature vectors to chasification, and just to the newly-increased workload classification , will not need all workload to lassificate again. And when algorithm in off-line, the system was finished automatically on the cluster based on feature vectors. when On-line for system, only need to classificate the newly-increased workload. So this measures save time and resource greatly .