Creating better datasets for weak scaling analysis.

[tinaok]

We have performed weak scaling analysis using chunk_size = 64, 128,256,512MB, starting 1 node.
we used chunk_per_worker 10 for all chunk size.
This creates non homogeneous data size for each analysis.
What about modifying this chunk_per_worker as 80,40,20 and 10 for chunk_size = 64, 128,256 and 512MB respectively?
Then the computational size is same for each chunk size, thus we should really be able to see the difference of chunk_size effect on computational time?