摘要
根据负载的变化,对数据库系统所提供的可调参数或选项进行手动调节是一项很困难的任务。把以概率推理为基础的图形化模型-影响图应用到数据库参数的自调节中,主要分为以下5个步骤:影响图模型的设计,实验数据集的获取,实验数据集的离散化,影响图参数的学习,影响图的评价及使用。以Oracle的重做日志模块的参数调节做实验,结果表明,影响图可以很好地解决这类问题。
Adjusting the parameters or ＂knobs＂ of DBMS manually in response to the changing workload is a difficult task. A probabilistic, graphical model known as influence diagram is introduced to adaptively adjust the parameters of DBMS. It divided into five steps： Designing the structure of the model, collecting the experimental data, discretizing the data, learning the model＇s parameters, evaluating and using the model. The experiment using this methodology to simultaneously adjust two knobs from redo log subsystem of Oracle showed that the influence diagram could effectively solve such problems.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第16期3812-3814,3817,共4页
Computer Engineering and Design
关键词
自调节
影响图
自调优
自管理
重做日志
adaptive adjustment
influence diagram
autonomic tuning
self-management
redo log
