数据库系统性能模型建模方法综述

数据库系统性能模型是数据库系统管理的重要基础技术支撑，广泛用于查询调度、资源分配、性能调优等任务中。当前的性能模型主要分为分析型和统计型两种，分析型模型需要深入研究数据库系统查询执行过程，对动态查询的适应性较好，无须成本高昂的采样实验，但在查询并行执行情景下建模复杂，对不同的数据库系统有不同的理论模型。统计型模型无须分析查询执行过程，通过采集查询执行参数并训练某个数学模型。统计型建模过程简单，能够较好地描述查询交互，预测效果较好，但采样成本很高，对动态查询的适应性差。对数据库系统性能建模的主要文献进行综述，重点介绍数据库系统性能建模的主要方法，并讨论这两类模型各自的优缺点、建模的难点以及应对策略。在此基础上，对数据库系统性能模型领域的研究做了展望，为有关该领域的研究提供参考。

Survey on performance modeling of database systems

Performance modeling of database systems plays an important role in managing database systems, which can be used for query scheduling, resource allocation, performance tuning, etc. The current performance models can be divided into two categories, analytic models and statistical models. The analytical models are built by studying the processes of query execution for the specific database system. It can adapt to dynamical queries well and does not need to do costly sampling experiments. It is complex, however, to describe the execution processes of concurrent queries, and has to develop different theoretical models for different database systems. In contrast, statistical models predict the performance of database systems by training a mathematical model with data collected from the database systems, avoiding the complexity of studying the detailed query execution processes and has better prediction accuracy. For the reason of sampling, statistical modeling is costly and cannot adapt to dynamic queries. This paper survives the literatures of performance modeling of database systems, puts emphasis on the major modeling methods, discusses the pros and cons of the two types of models, and the challenging and corresponding approaches for attacking them. Finally, the future research directions in the field of database performance modeling are discussed.