REPORTING
There are two ways to generate a performance report, at runtime and post-mortem.
RUNTIME REPORTING
The hydra master controller will, upon request, autogenerate a performance report at the end of a test run. This is controlled via the configuration parameter:
perffmwk.PerfReportPrms-generatePerformanceReport = <boolean>;
The report is written to "perfreport.txt" in the test directory. It is based on statistics and trim specification files autogenerated by the framework.
The autogenerated statistics specification file is written to "statistics.spec" in the test directory. By default, it contains default specifications for any user-defined statistics created by the test code at runtime, along with a pointer to the default specification file $JTESTS/perffmwk/statistics.spec containing common system statistics. A test can optionally specify its own statistics specification file, using:
perffmwk.PerfReportPrms-statisticsSpecification = <filename>;
In this case, the autogenerated "statistics.spec" file will simply contain a pointer to the specified file. See statspec_grammar.txt for details on how to construct a statistics specification file. Tokens for the grammar are defined in {@link perffmwk.StatSpecTokens}.
The autogenerated trim specification file is written to "trim.spec" in the test directory. It contains endpoints for named trim intervals specified by the test code at runtime, if any. Trim intervals with default endpoints are omitted.
POST_MORTEM REPORTING
The post-mortem reporting tool has the system property "perffmwk.statisticsSpecification", which defaults to "statistics.spec" in the test directory. The trim specification used is always "trim.spec" in the test directory.
The statistics and trim specification files in individual test directories can also be modified in place before invoking post-mortem tool.
See statspec_grammar.txt and trimspec_grammar.txt for details on how to construct specifications.
For how to run the post-mortem performance reporter, see {@link perffmwk.PerfReporter}.