Create a performance comparison between running JobQueue and Scheduler

[xjules]

This should include running the following experiments on rgs node(s):

• [x] big poly example
• [x] drogon (test_run)
• [x] default poly example

[JHolba]

all tested with local driver. the RGS node had 6 cores.

bigpoly rgs 16m13s "ert es_mda --disable-scheduler bigpoly.ert" 15m31s "ert es_mda --enable-scheduler bigpoly.ert"

laptop ❯ hyperfine --prepare "rm -rf poly_out storage logs" "ert es_mda --enable-scheduler bigpoly.ert" "ert es_mda --disable-scheduler bigpoly.ert" Benchmark 1: ert es_mda --enable-scheduler bigpoly.ert Time (mean ± σ): 582.667 s ± 12.119 s [User: 3581.944 s, System: 1015.406 s] Range (min … max): 570.126 s … 607.157 s 10 runs

Benchmrak 2: ert es_mda --disable-scheduler bigpoly.ert Time (mean ± σ): 631.161 s ± 24.146 s [User: 3770.351 s, System: 1129.080 s] Range (min … max): 607.496 s … 670.939 s 10 runs

Summary ert es_mda --enable-scheduler bigpoly.ert ran 1.08 ± 0.05 times faster than ert es_mda --disable-scheduler bigpoly.ert

poly example rgs

disabled real 1m16.837s user 1m42.647s sys 0m42.980s

enabled real 1m7.036s user 1m17.980s sys 0m32.200s

laptop ❯ hyperfine --prepare "rm -rf poly_out storage logs" "ert es_mda --enable-scheduler poly.ert" "ert es_mda --disable-scheduler poly.ert" Benchmark 1: ert es_mda --enable-scheduler poly.ert Time (mean ± σ): 35.843 s ± 0.307 s [User: 132.677 s, System: 29.183 s] Range (min … max): 35.542 s … 36.639 s 10 runs

Benchmark 2: ert es_mda --disable-scheduler poly.ert Time (mean ± σ): 51.088 s ± 1.710 s [User: 170.195 s, System: 23.729 s] Range (min … max): 49.319 s … 54.804 s 10 runs

Summary ert es_mda --enable-scheduler poly.ert ran 1.43 ± 0.05 times faster than ert es_mda --disable-scheduler poly.ert

[JHolba]

drogon rgs 11m30s --enabled 12m4s --disabled

[pinkwah]

Closing as a successful performance comparison. Well done!