Running RAxML Efficiently On a Cluster

[lsa...@aucklanduni.ac.nz]

Hi,

I recently ran RAxML on 3 data sets with one run finishing in 27 hours and the other 2 timing out after a week, the smaller set managed to complete the bootstrapping and had started the ML search however. I was wondering If I could get some advice for running it more quickly before I requeue.

Both datasets have 68 taxa and one has 2,846,057 sites and the other has 11,621,084. My phylip files contain the invariant sites so the actual number of SNPs are 277,075 and 1,010,050 respectively.

I ran raxmlHPC-SSE3 with -f a, GTRGAMMA and -N autoMR the RAxML manual says -T is only for the pthreads versions.

I submit the jobs to cluster with slurm files where I used the following options:

#SBATCH --ntasks=4

#SBATCH --mem-per-cpu=2g

#SBATCH --cpus-per-task=4

The cluster I'm using is here: https://wiki.auckland.ac.nz/display/CER/NeSI+Pan+Cluster

What could I be doing better to optimize it before I queue these runs up again? Should I run a different version of RAxML? Any help would be greatly appreciated.

Liam

[Alexandros Stamatakis]

Dear Liam,

For datasets as large as these you should better run our ExaML code
which has been specifically designed for analyzing such large dataset.

As a rule of thumb you should use one core to work on 1000 alignment sites.

You can also use RAxML-NG which uses the same parallelization approach
as ExaML which is much faster than what is implemented in RAxML for such
long datasets with millions of sites.

Alexis

> --
> 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
> an email to raxml+un...@googlegroups.com
> <mailto:raxml+un...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Alexandros (Alexis) Stamatakis

Research Group Leader, Heidelberg Institute for Theoretical Studies
Full Professor, Dept. of Informatics, Karlsruhe Institute of Technology
Adjunct Professor, Dept. of Ecology and Evolutionary Biology, University
of Arizona at Tucson

www.exelixis-lab.org