gmx_MMPBSA multiple frame analysis

[Nidhin Thomas]

Hi everyone,

I have successfully analyzed multiple frames of a membrane-protein-ligand complex with gmx_MMPBSA. Size of my system is roughly 7 X 7 X 8 nm with ~45000 particles. 

I am able to analyze about 11 frames of my system with 11 processors within 1 hour. However, the problem arises when I try to perform multiple analyses with different systems simultaneously. My speed reduces drastically and for two separate system, it takes ~2 hours per system. I have attached my relevant slurm bash script below. I have looked into the cpu allocation and it seems there is no overlap between these two systems. 

I might be missing something here. Can anyone please help me understand where I may be going wrong? My expectation is that I should be able to analyze multiple systems as long as I have CPUs available.

#!/bin/bash

#SBATCH -J MMPBSA

#SBATCH -o MMPBSA.o%j

#SBATCH -t 12:00:00

#SBATCH -N 1 -n 11

#SBATCH --ntasks-per-core 1

module load ambertools

module load gromacs/2021.2

module load gmx_MMPBSA

 mpirun -np 11 gmx_MMPBSA MPI  -O -i mmpbsa.in -cs system.gro -ci index.ndx -cg 4 5 -ct npt.trr -cp topol.top -nogui

sleep 5s

scontrol status:

scontrol show job -d 1016

JobId=1016 JobName=MMPBSA

   UserId=******* GroupId=platform(502) MCS_label=N/A

   Priority=1260 Nice=0 Account=(null) QOS=normal

   JobState=RUNNING Reason=None Dependency=(null)

   Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0

   DerivedExitCode=7:0

   RunTime=1-04:40:48 TimeLimit=3-00:00:00 TimeMin=N/A

   SubmitTime=2021-11-07T15:33:47 EligibleTime=2021-11-07T15:33:47

   AccrueTime=2021-11-07T15:33:47

   StartTime=2021-11-07T15:33:48 EndTime=2021-11-10T15:33:48 Deadline=N/A

   SuspendTime=None SecsPreSuspend=0 LastSchedEval=2021-11-07T15:33:48 Scheduler=Main

   Partition=gpu-dev AllocNode:Sid=gpu-1:939333

   ReqNodeList=(null) ExcNodeList=(null)

   NodeList=gpu-2

   BatchHost=gpu-2

   NumNodes=1 NumCPUs=22 NumTasks=11 CPUs/Task=1 ReqB:S:C:T=0:0:*:*

   TRES=cpu=22,node=1,billing=22

   Socks/Node=* NtasksPerN:B:S:C=0:0:*:1 CoreSpec=*

   JOB_GRES=(null)

     Nodes=gpu-2 CPU_IDs=0-11,128-137 Mem=0 GRES=

   MinCPUsNode=1 MinMemoryNode=0 MinTmpDiskNode=0

   Features=(null) DelayBoot=00:00:00

   OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)

I also tried without --ntask-per-core 1

scontrol show job -d 1022

JobId=1022 JobName=MMPBSA

   UserId=******** GroupId=platform(502) MCS_label=N/A

   Priority=1260 Nice=0 Account=(null) QOS=normal

   JobState=RUNNING Reason=None Dependency=(null)

   Requeue=1 Restarts=0 BatchFlag=1 Reboot=0 ExitCode=0:0

   DerivedExitCode=0:0

   RunTime=00:00:25 TimeLimit=12:00:00 TimeMin=N/A

   SubmitTime=2021-11-08T20:13:04 EligibleTime=2021-11-08T20:13:04

   AccrueTime=2021-11-08T20:13:04

   StartTime=2021-11-08T20:13:05 EndTime=2021-11-09T08:13:05 Deadline=N/A

   SuspendTime=None SecsPreSuspend=0 LastSchedEval=2021-11-08T20:13:05 Scheduler=Main

   Partition=gpu-dev AllocNode:Sid=gpu-2:1035731

   ReqNodeList=(null) ExcNodeList=(null)

   NodeList=gpu-2

   BatchHost=gpu-2

   NumNodes=1 NumCPUs=12 NumTasks=11 CPUs/Task=1 ReqB:S:C:T=0:0:*:*

   TRES=cpu=12,node=1,billing=12

   Socks/Node=* NtasksPerN:B:S:C=0:0:*:* CoreSpec=*

   JOB_GRES=(null)

     Nodes=gpu-2 CPU_IDs=12-17,138-143 Mem=0 GRES=

   MinCPUsNode=1 MinMemoryNode=0 MinTmpDiskNode=0

   Features=(null) DelayBoot=00:00:00

Any help would be appreciated!

Thanks a lot,

Nidhin Thomas

[marioe911116]

Hello Nidhin and welcome! 

I wonder what does this mean ...when I try to perform multiple analyses with different systems simultaneously

Is it that you are running different analyses for different systems using several slurm scripts (one per system)?

could you please describe in detail your procedure and include the relevant slurm scripts as well as the mmpbsa.in files?

cheers!

Mario E.