Speed Issue when using a university Cluster

[Rachel Isaac]

I am running an alignment of about 6,000 sequences with a length of 230 amino acids. I am wanting to build a maximum likelihood tree. When I run it on my local machine, it takes up a lot of computing power and makes everything else run slower so I am trying to run it on my university's cluster (Negishi). When I test it with a small alignment (25 sequences), it takes 6 seconds on my local machine and 24 seconds on the cluster. Does anyone have any ideas?

This is the script I use for the local computer:

raxml-ng --msa Round31.fasta --model LG4X --prefix Round31

This is my bash script:

#!/bin/bash
#SBATCH -A bphm
#SBATCH -t 336:00:00
#SBATCH -N 1
#SBATCH --ntasks=4
#SBATCH --cpus-per-task=8
#SBATCH --job-name=raxml-ng
#SBATCH --mail-type=FAIL,BEGIN,END
#SBATCH --error=/scratch/negishi/isaac9/outputs/round31/%x-%J.err
#SBATCH --output=/scratch/negishi/isaac9/outputs/round31/%x-%J.out

module --force purge
ml biocontainers raxml-ng

raxml-ng --msa /scratch/negishi/isaac9/alignments/round31.fasta --model LG4X --prefix /scratch/negishi/isaac9/outputs/round31/round31

[Oleksiy Kozlov]

Hi Rachel,

since you are apparently not using MPI, you should set

#SBATCH --ntasks=1
#SBATCH --hint=compute_bound

Also, please attach full raxml-ng log, then I will see what is going on.

Best,
Oleksiy

On 14.05.25 15:12, Rachel Isaac wrote:
> I am running an alignment of about 6,000 sequences with a length of 230 amino acids. I am wanting to
> build a maximum likelihood tree. When I run it on my local machine, it takes up a lot of computing

> power and makes everything else run slower so I am trying to run it on myuniversity's cluster
> (Negishi). When I test it with a small alignment (25 sequences), it takes6 seconds on my local

> machine and 24 seconds on the cluster. Does anyone have any ideas?
>
> This is the script I use for the local computer:
>
> raxml-ng --msa Round31.fasta --model LG4X --prefix Round31
>
> This is my bash script:
>
> #!/bin/bash
> #SBATCH -A bphm
> #SBATCH -t 336:00:00
> #SBATCH -N 1
> #SBATCH --ntasks=4
> #SBATCH --cpus-per-task=8
> #SBATCH --job-name=raxml-ng
> #SBATCH --mail-type=FAIL,BEGIN,END
> #SBATCH --error=/scratch/negishi/isaac9/outputs/round31/%x-%J.err
> #SBATCH --output=/scratch/negishi/isaac9/outputs/round31/%x-%J.out
>
> module --force purge
> ml biocontainers raxml-ng
>
> raxml-ng --msa /scratch/negishi/isaac9/alignments/round31.fasta --model LG4X --prefix /scratch/
> negishi/isaac9/outputs/round31/round31
>

> --
> You received this message because you are subscribed to the Google Groups"raxml" group.
> To unsubscribe from this group and stop receiving emails from it, send anemail to
> raxml+un...@googlegroups.com <mailto:raxml+un...@googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/raxml/41200722-72d2-45a7-9b8c-
> e2eebfb40836n%40googlegroups.com <https://groups.google.com/d/msgid/raxml/41200722-72d2-45a7-9b8c-
> e2eebfb40836n%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
Dr. Oleksiy Kozlov
Staff scientist, Sustainable Computing Advisor

Computational Molecular Evolution (CME)
Heidelberg Institute of Theoretical Studies (HITS gGmbH)
Schloss-Wolfsbrunnenweg 35, 69118 Heidelberg, Germany

Amtsgericht Mannheim / HRB 337446
Managing Director: Dr. Gesa Schönberger
Scientific Director: Dr. Kai Polsterer

[Oleksiy Kozlov]

Hi Rachel,

since you are apparently not using MPI, you should set

#SBATCH --ntasks=1
#SBATCH --hint=compute_bound

Also, please attach full raxml-ng log, then I will see what is going on.

Best,
Oleksiy

On 14.05.25 15:12, Rachel Isaac wrote:

> I am running an alignment of about 6,000 sequences with a length of 230 amino acids. I am wanting to
> build a maximum likelihood tree. When I run it on my local machine, it takes up a lot of computing

> power and makes everything else run slower so I am trying to run it on myuniversity's cluster
> (Negishi). When I test it with a small alignment (25 sequences), it takes6 seconds on my local

> machine and 24 seconds on the cluster. Does anyone have any ideas?
>
> This is the script I use for the local computer:
>
> raxml-ng --msa Round31.fasta --model LG4X --prefix Round31
>
> This is my bash script:
>
> #!/bin/bash
> #SBATCH -A bphm
> #SBATCH -t 336:00:00
> #SBATCH -N 1
> #SBATCH --ntasks=4
> #SBATCH --cpus-per-task=8
> #SBATCH --job-name=raxml-ng
> #SBATCH --mail-type=FAIL,BEGIN,END
> #SBATCH --error=/scratch/negishi/isaac9/outputs/round31/%x-%J.err
> #SBATCH --output=/scratch/negishi/isaac9/outputs/round31/%x-%J.out
>
> module --force purge
> ml biocontainers raxml-ng
>