search

tbarbette / npf

Network Performance Framework: easy-to-use experiment manager with automated testing, result collection, and graphing
GNU General Public License v3.0
38 stars 15 forks source link

Support for CDFs graph #21

Open tbarbette opened 2 years ago

tbarbette commented 2 years ago

Following #13, a native support for CDFs would be nice.

The idea is that all results for each run would be printed as a line for CDFs.

Either a single test produce many RESULT-XXX in stdout, so we have many points and can draw a CDF showing the percentage of points in Y, and the value range in X from that.

Or, one have "noise" variables, and take all results from the noise in a single variable using var_aggregate={VAR1+VAR2+VAR3:all}.

Maybe I should add a prefix to variables to automatically "hide" them, i.e one would be able to add a variable such as @LOSS=[0-10], showing we don't really care about the result for a particular LOSS rate but want to re-run tests for many LOSS values and its impact on the variance (through a CDF, or errorbars, or boxplot).

tbarbette commented 2 years ago

Now supported, see https://github.com/tbarbette/npf/blob/4be5cf74a01c05226e0c4f30ce95e0739e9ebe50/integration/cdf.npf . But I still need to write the doc