Described in the blog post.
To reproduce
- Run
make benchmark
, which will generatebenchmark.log
- Run
bash generate_histograms.sh
to generate histograms for node access times. - Run
jupyter lab
and runplot_benchmark.ipynb
to get the table with benchmark timings. - Run
plot_histograms.ipynb
to plot the "cache friendlieness" histograms. - Run
generate_rebalance_counters.sh
to generate rebalance counts. - Run
plot_rebalance_cnt.ipynb
to plot the counts.