The performance trace facility is a facility that collects trace information for troubleshooting, which is output in response to a series of HiRDB processes. To make it easier to identify the process that has caused a problem, the performance trace facility also collects trace information on processes that tend to result in performance bottlenecks, such as disk I/O operations and process-to-process communication (transmission and receiving). The facility can also collect trace information on processing that occurs among multiple machines or multiple processes.

The statistical information used for conventional troubleshooting is meant to assist in checking the operating status of the entire HiRDB system. In contrast, performance trace information can be used to check the operating status of each connection. By sending information to the support service, you can more efficiently sort delay factors and identify problem causes, leading to faster resolutions. Furthermore, if you edit the output trace information and output it to a CSV file, the information can also be analyzed in a more flexible manner using a tool that can handle CSV format, such as Excel.

Note

When performance trace information is collected, since concurrent execution of processes is given higher priority, no locking is applied. As a result, some trace information might not be output or, in some cases, invalid information might be output. For details about how to make sure that the performance trace information you are referencing is normal trace information, see 28.3 Notes.

The following figure provides an overview of the performance trace facility.

Figure 28-1 Overview of the performance trace facility (when used for a HiRDB parallel server configuration)

Explanation

1. During HiRDB processing, performance trace information is collected at collection points. This performance trace information is output automatically to the performance trace information file at 10-second intervals. At this stage, the performance trace information is in binary format.
2. Execute the pdprfed command to edit and convert the performance trace information to CSV format.
3. Using a tool that can handle CSV files, analyze the problem cause.

Organization of this section

28.1.1 Processes targeted by the performance trace facility

28.1.2 Settings for collecting performance trace information

28.1.3 Using the performance trace information file

28.1.4 Acquiring performance trace information when linked with Hitachi Application Server