We need to go to something as high as 2 million processes to check that we still have the same performance than in previous article.

You only give the speedup achieved by the parallelization, but I would like to also see the absolute values. How many hours are we spending in these simulations?

The Figure "Proportions of SR’s having only one process to compute" would be easier to read if it could be depicted as a stacked bar chart. See http://stackoverflow.com/questions/20349929/stacked-bar-plot-in-r

Also, I'd like to see more categories, like 1 process, 2 processes, 3-5 processes, 6-10, 11-20, 21-30, 31+
That's important because in section 5.1, you say that parallel execution seems interesting when given more than 28 processes, so it'd be good to track the amount of corresponding SRs. Or maybe I'm misunderstanding something?

I know that memory consumption is not something that we will improve through parallelization, but it would still be interesting to have some graphs about it. That could be placed in the section studying the v3.11 baseline performance.