Table 3 Performance of docking of 5000 ligands to CDK2 (2BTR) with Autodock Vina using different number of computational nodes
Number of computational nodes (parallelization) | Total number of cores | N workers per node | N cpu per molecule | Wall time | Speed up |
|---|---|---|---|---|---|
1 (multiprocessing, random priority) | 32 | 8 | 5 | 7Â h 4Â m | 1 |
1 (dask) | 32 | 8 | 5 | 7Â h 19Â m | 0.966 |
2 (dask) | 64 | 8 | 5 | 3Â h 39Â m | 1.936 |
5 (dask) | 160 | 8 | 5 | 87Â m 43Â s | 4.833 |
10 (dask) | 320 | 8 | 5 | 44Â m 8Â s | 9.607 |
20 (dask, random priority) | 640 | 32 | 1 | 29Â m 37Â s | 14.32 |
20 (dask) | 640 | 32 | 1 | 26Â m 45Â s | 15.85 |
20 (dask) | 640 | 16 | 2 | 23Â m 43Â s | 17.88 |
20 (dask) | 640 | 16 | 3 | 23Â m 21Â s | 18.16 |
20 (dask) | 640 | 8 | 4 | 22Â m 19Â s | 19.00 |
20 (dask) | 640 | 8 | 5 | 22Â m 14Â s | 19.07 |
20 (dask, random priority) | 640 | 8 | 5 | 22Â m 35Â s | 18.77 |