Evolution of the docking process

[bureau]

The speed of screening strongly depends on the number of compounds waiting. I find it difficult to understand this phenomenon and particularly at the end of the process for which the speed decreases strongly.
I start with 80 millions compound with around 10000 cpus and 41 jobfiles (10 (step) *24 for each jobline). The ligand-todo-per-queue is 5000 and the ligand-per-refilling_step is 1000.
Could you give me an explanation ?
Ronan,

[Christoph]

Hi Ronan,

Could you explain in more detail what you mean with “number of compounds waiting”? And how do you measure the speed of screening?

Here is also another post which elaborates on the scaling behavior and variance int he completion time:

Scaling question

I managed to get set up on GCP with SLURM as a job manager. I am trying to see how well the scaling works. I have been using your tutorial 1 example as a test case. 1200 or so ligands there against GK (I think). I ran a 15 cpu job and finished it in 3 hours 45 min. I tried a 95 cpu job and finished in ~1 hour. So 6x cpu only brought me <4x increase. all.ctrl has following settings: steps_per_job=1 cpus_per_step=95 queues_per_step=95 cpus_per_queue=1 I changed some of the ligand filli…

Best,
Christoph

[bureau]

Hi Christoph,

The first run is correct (associated with the first queues). Starting from 41 joblines and 10 steps by jobline, 24 CPUs, all start correctly. I have around 5000 compounds by queues. The problem happens for the second run. Some joblines start and others wait. In those waiting, few applications of qvina2 (in a strange behavior, running / stopping …) still run (among the 240 CPUs associated to the jobline). The best way to maintain the level of VS, at this level, is currently to stop the overall process and run again.

I try to find some keys to understand the problem.

Best,

Ronan