TOM v0.1¶
TOM is a test tesults analyzer and benchmarker. Its features are: -Benchmarker: configuration comparator --rank configurations according to median test result value --Visualize configuration differences using boxplots -Analyzer: determine how fare a set of configuration parameters impacts the result of test --evaluation of the predictability of test result --according to each parameter --according to a combination of parameters --algorithm of to find the best combination
Some algorithmic aspects¶
The tool processes the data to evaluate the explanatory power of the constituting factors (test context variables) regarding the key point indicator (KPI) of the test. The coefficient of determination is used for this purpose. It is calculated as follows:
Analysis examples¶
Best subset of context variables to explain KPI of interest ('bandwidth.MBps.')¶
Explaination power of each subset of variables¶
Benchmarking examples¶
Benchmarking deployment scenarios according to the obtained memory bandwidth test results¶
Benchmarking PODs according to the obtained memory bandwidth test results¶
See OPNFV online wiki page for more usage examples.
Original author¶
Alassane Samba, Orange
For future version¶
- Enable API parameter pass using json (data file and and variable parameters)