Figure 14
In this experiment, we validate Musketeer's automated mapping choices for six workflows in 33 setups by comparing them against the correct choices and recording the degradation in makespan when Musketeer fails to identify the best system. Without workflow-specific information, Musketeer only makes choices within 10% of the best choice's makespan in around half the cases. However, with workflow-specific information (input, output and intermediate data sizes), the automated mapping improves. With information about all operators in a workflow, Musketeer chooses an option with a makespan no more than 10% from the best choice.
Figure 14
Under construction: we will add information on the experimental setup and our data sets here shortly.
If you are interested in being notified when the data appears,
please join our
musketeer-announce mailing list.
Thanks for your patience.
-- The Musketeer team.
Experimental setup
This experiment was executed on the freestyle machine from our small dedicated cluster of seven machines.
Result data set
The raw results for this experiment are available here.
To plot Figure 14, run the following command:
experiments/plotting_scripts$ python plot_scheduler_experiment.py ../scheduler/scheduling_eval.csv "Decision tree" "First run" "+ partial history" "+ full history" scheduler_experiment.pdf
The graph will be in scheduler_experiment.pdf