|
About | News | Download | Help | nmichael.de |
Version 2.1.1 adds the ability to temporarily hide/unhide series from the graph to help exploration of large sets of series. It also adds support for CSV file merging and conversion as well as data correlation on the command line. Detection of timestamps has been improved. Several bugfixes and minor improvements for the hotspot, iostat, csv, and nicstat parsers.
Version 2.1.0 adds a simple reporter to PDA used for command-line report and project generation as well as improved statistics selection dialogs. Small enhancements include improvements and bugfixes to existing parsers as well as the addition of a parser for Linux "perf" command.
In version 2.0.5, PDA now supports external converter plugins that transform an input file, parsed by a PDA parser, to another format. Additionally, some parsers have been improved and enhanced to support further formats.
PDA version 2.0.4 adds support for command-line plotting, which enables users to plot arbitrary series for selective files directly from the command line. For existing projects, PDA can now also replace the path of all files in a project with another parse, a useful feature when the same data (same file names) have been collected multiple times (in different directories) and shall be evaluated with PDA in the same way for each test.
Version 2.0.2 brings a few minor enhancements to PDA: Users can now for example configure the size and colors of labels (which now also appear on the highest layer and are no longer sitting behind the plotted curves) and also have a simple way to add series names as additional labels to the graph. Also the default sizes of the labels, including those used for the legend, can now be configured, and their position is determined based on their actual needs, thus avoiding overlapping labels in case of long series names. A new API allows parsers to provide labels for any data points which are automatically placed above the respective points in the graph.
The new version 2.0.1 improves the zoom handling of PDA: Instead of forcing users to reset the zoom completely after zooming in, PDA now allows to flexibly zoom in and out, and scroll left and right while maintaining the current zoom settings.
Performance Data Analyzer (PDA) is a tool to analyze performance data and troubleshoot problems by correlating data from various sources. It comes with parsers for many common OS, Java, and database statistics, including parsers for vmstat, mpstat, iostat, collectl, Java garbage collection, and Oracle database statistics. PDA transfers data from these sources into internal time series. These series can then be plotted in a common graph, aligned to a common time axis and scaled individually. By interactively zooming in and out, different time periods can be analyzed in depth to identify or root-cause problems -- for example one curve spiking while another one drops. Series can be dynamically added and removed from the graph, grouped with other series, plotted in different styles, smoothed and scaled based on needs. The statistics function of PDA allows to calculate statistics such as minimum, maximum, average, variance or distance between samples for any of these data series, either for the entire time or any selected interval. The pluggable parser framework of PDA allows to easily add new parsers for other data formats.
Today I am releasing the partially rewritten version 2 of PDA. It comes with a more flexible grouping of series, which even allows to group them across parsers, simplified configuration and selection of series through filtering, an improved user interface, enhanced parser framework, and reduced memory consumption for projects with lots of data. PDA is available under the GNU General Public License v2 and can be downloaded at pda.nmichael.de.
Nicolas Michael
| PDA - Performance Data Analyzer • Copyright © 2006-18 by Nicolas Michael • Contact |