The constructDuration.py is the slowest one. Strangely it is slow also at experiment level, although I expect it to be really fast given that it simply takes already computed metrics from the trial metrics.
For the trial level evaluate the following improvements:
- Reuse the
filteredRDD instead of basically computing it again, and also reuse the data variable where you basically have all the information you compute again for the for
The
constructDuration.pyis the slowest one. Strangely it is slow also atexperimentlevel, although I expect it to be really fast given that it simply takes already computed metrics from thetrialmetrics.For the
triallevel evaluate the following improvements:filteredRDDinstead of basically computing it again, and also reuse thedatavariable where you basically have all the information you compute again for thefor