云环境下基于LQNM的数据库性能建模研究

STUDY ON DATABASE PERFORMANCE MODELLING BASED ON LQNM IN CLOUD

在云环境下,数据库系统性能的优劣是云服务供需双方共同关注的焦点,如何形象地描述云环境下数据库系统的资源分配,合理地评价数据库系统的性能成为目前一个重要的问题。通过分析云环境下数据库系统的工作原理,提出采用分层排队网络建模技术为其建立性能模型,该模型把云环境下的数据库系统抽象为三个软件任务(客户端、查询代理进程、数据管理进程)和三个硬件任务(THINK、CPU、硬盘),并以负载特征参数作为输入,通过分层的方法求出该模型的解。通过模型预测值与实验测量值分析系统平均响应时间之间的误差,并与传统的排队网络模型进行比较,从而验证该性能模型的有效性和准确性。

In cloud environment, the advantages and disadvantages of database system＇ s performance are the common focus concerned by both the cloud providers and the users, how to vividly describe the resources allocation of database system in cloud and reasonably evaluate the performance of database system have become an important issue at present. We propose to use layered queuing network modelling （LQNM） technique to build the performance model of database system by analysing the working principle of it in cloud, this model abstracts the database system in cloud environment as three software tasks, namely, the client, the query agent process and the data management process, and three hard tasks, the THINK, CPU and DISK. With the workload feature parameters as the inputs, we obtain the solution of model through layered method. The validity and accuracy of the performance model are verified through analysing the error between average system responding times with model prediction value and measured value in experiment, and comparing with traditional queuing network models.