A performance management methodology for oracle RDBMS on HP-UX

摘要

The basic methodology for performance management of a Relational Database Management System (RDBMS) involves steps like assessment, measurement, interpretation and analysis, identification of bottlenecks and tuning. Performance of a relational database depends upon the features and capabilities of the operating system. So it is the joint task of any database and system vendor to appropriately manage their respective features in order to achieve the desired level of performance. This paper first describes the above steps involved in a performance management methodology. It talks about some of the RDBMS performance features that one needs to understand and properly make use of in order to enhance the overall server performance. During the course of the discussion, it also refers to the common performance bottlenecks and standard tools to detect and tune each of these bottlenecks at the system and database level. As operating system and database level tuning are not always enough, one technique for improving performance is to create efficient machine code and optimal programs by optimizing at the compiler and linker levels. This paper mentions the importance of proper utilization of the platform-specific compiler and linker flags for building optimal executables. In addition to the above, designing and profiling the application for optimality also enhances performance in general. This paper mentions the advantage of Profile-based or similar optimization. It also recommends configuring the system kernel parameters and granting necessary privileges best suited for an optimal execution of the RDBMS. This paper also mentions the importance of proper use of asynchronous I/O, process scheduling policy, etc. From time to time this paper will draw examples from Oracle RDBMS running on HP-UX. To summarize, this paper discusses a performance management methodology towards a well-tuned system with an optimal relational database and an efficient application running against it.