beast 2.1.3 initializing run takes more than 24 hours. why?

[MarianneE]

Hi,

I have been running BEAST 2.1.3 on my macbook (from command line) and on the CIPRES cluster to do divergence dating on a dataset with 5 loci (3666bp, partitioned into codon positions with unlinked substitution models, tree prior set to birth-death, four calibration points) and 201 taxa (50 million generations, sampling every 5000 as a first test). In both instances it takes more than 30 hours before the Markov chain actually starts after the analysis is up and running. It seems that it for some reason takes a really long time to write the tree file (see output below, this is from CIPRES). There is no further output during these ~30 hours of apparently writing the tree file. Does anyone know why this takes so long and what might be happening? I previously ran the same data using BEAST 1.8.5 and a similar run started within seconds and finished in less than four days. 

One my macbook I had to use the -java flag to avoid Beagle/CUDA issues, which I know slows everything down, but this shouldn't be the issue since the same thing happens on CIPRES.

Thanks,

Marianne

CIPRES_THREADSPP=8

CIPRES_NP=1

NOTICE : Your input file may have been altered to activate multi-threading.

running:

java -Xmx15000m -Djava.library.path=/projects/ps-ngbt/opt/gordon/beast2/2.1.3/lib:/opt/beagle/2.1/lib -cp /projects/ps-ngbt/opt/gordon/beast2/2.1.3/lib/beast.jar:/projects/ps-ngbt/opt/gordon/beast2/2.1.3/BEASTlabs/lib/BEASTlabs.addon.jar:/projects/ps-ngbt/opt/gordon/beast2/2.1.3/BEAST_CLASSIC/lib/BEAST_CLASSIC.addon.jar beast.app.beastapp.BeastMain -seed 12345 -overwrite -beagle_scaling dynamic infile_altered.xml

                  BEAST v2.1.3, 2002-2014

       Bayesian Evolutionary Analysis Sampling Trees

                 Designed and developed by

Remco Bouckaert, Alexei J. Drummond, Andrew Rambaut and Marc A. Suchard

                              

               Department of Computer Science

                   University of Auckland

                  re...@cs.auckland.ac.nz

                  ale...@cs.auckland.ac.nz

                              

             Institute of Evolutionary Biology

                  University of Edinburgh

                     a.ra...@ed.ac.uk

                              

              David Geffen School of Medicine

           University of California, Los Angeles

                     msuc...@ucla.edu

                              

                Downloads, Help & Resources:

              http://beast2.cs.auckland.ac.nz

                              

Source code distributed under the GNU Lesser General Public License:

              http://code.google.com/p/beast2

                              

                     BEAST developers:

Alex Alekseyenko, Trevor Bedford, Erik Bloomquist, Joseph Heled, 

Sebastian Hoehna, Denise Kuehnert, Philippe Lemey, Wai Lok Sibon Li, 

Gerton Lunter, Sidney Markowitz, Vladimir Minin, Michael Defoin Platel, 

          Oliver Pybus, Chieh-Hsi Wu, Walter Xie

                              

                         Thanks to:

    Roald Forsberg, Beth Shapiro and Korbinian Strimmer

Random number seed: 12345

201 taxa

3666 sites

2338 patterns

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

  Using BEAGLE version: 2.1 resource 0: CPU

    with instance flags:  PRECISION_DOUBLE COMPUTATION_SYNCH EIGEN_REAL SCALING_MANUAL SCALERS_RAW VECTOR_NONE THREADING_NONE PROCESSOR_CPU FRAMEWORK_CPU

======================================================

Please cite the following when publishing this model:

Remco Bouckaert, Joseph Heled, Denise Kuehnert, Tim Vaughan, Chieh-Hsi Wu, Dong Xie, Marc Suchard, Andrew Rambaut, Alexei J Drummond BEAST 2: A software platform for Bayesian evolutionary analysis. PLOS Computational Biology 10(4): e1003537, 2014

Gernhard 2008. The conditioned reconstructed process. Journal of Theoretical Biology Volume 253, Issue 4, 21 August 2008, Pages 769-778

Drummond AJ, Ho SYW, Phillips MJ, Rambaut A (2006) Relaxed Phylogenetics and Dating with Confidence. PLoS Biol 4(5): e88

Hasegawa, M., Kishino, H and Yano, T. 1985. Dating the human-ape splitting by a molecular clock of mitochondrial DNA. Journal of Molecular Evolution 22:160-174.

======================================================

Writing file all201.log

         Sample      posterior ESS(posterior)     likelihood          prior

Writing file tree.trees

[Remco Bouckaert]

Hi Marianne,

Is it possible that you specified pre-burnin to be a large number for this analysis? During pre-burnin, the MCMC runs, but no samples are logged, and there is no  output to screen during that time.

Cheers,

Remco

[MarianneE]

Hi Remco,

Thanks for replying!

No, pre-burnin was set to the first 5 mill. generations. Also, output to the screen started immediately in 1.8.5 with same settings and I guess there shouldn't be that large differences between versions?

Thanks,

Marianne

[Andrew Rambaut]

BEAST 1.8 doesn’t have a ‘pre-burnin’ option. Try setting the pre-burnin to 0 and see if it starts straight away.

Andrew

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To post to this group, send email to beast...@googlegroups.com.
Visit this group at http://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.

Andrew Rambaut 

Institute for Evolutionary Biology | Centre for Infection, Immunity & Evolution 

Ashworth Laboratories, University of Edinburgh, Edinburgh, EH3 9JT, UK

contact – a.ra...@ed.ac.uk | http://tree.bio.ed.ac.uk | tel +44 131 6508624

[MarianneE]

Hi Andrew,

Thanks! Yes that did it. With preburnin=0 it starts right away. But why is Beast 2 so much slower than 1.8? For 1.8 it takes around 0.45 h/mill. generations while in 2.1.3 it takes almost two hours/mill gens, with as similar settings as possible.

Thanks,

Marianne

[Andrew Rambaut]

Check in the initial text that comes up that 2.1.3 is using BEAGLE.

Andrew

[MarianneE]

BEAST 2.1.3 is using beagle (see below) so that is not the issue. 

Marianne

CIPRES_THREADSPP=8
CIPRES_NP=1
NOTICE : Your input file may have been altered to activate multi-threading.
running:
java -Xmx15000m -Djava.library.path=/projects/ps-ngbt/opt/gordon/beast2/2.1.3/lib:/opt/beagle/2.1/lib -cp /projects/ps-ngbt/opt/gordon/beast2/2.1.3/lib/beast.jar:/projects/ps-ngbt/opt/gordon/beast2/2.1.3/BEASTlabs/lib/BEASTlabs.addon.jar:/projects/ps-ngbt/opt/gordon/beast2/2.1.3/BEAST_CLASSIC/lib/BEAST_CLASSIC.addon.jar beast.app.beastapp.BeastMain -seed 12345 -beagle_scaling dynamic infile_altered.xml

                  BEAST v2.1.3, 2002-2014
       Bayesian Evolutionary Analysis Sampling Trees
                 Designed and developed by
Remco Bouckaert, Alexei J. Drummond, Andrew Rambaut and Marc A. Suchard
                              
               Department of Computer Science
                   University of Auckland
                  

[Miller, Mark]

As an update to this issue, I wanted to say that on investigation we found an error in our CIPRES implementation of BEAST2, and this caused the slowness we (and Marianne) saw with her data set.

 

After repairing the issue, we found that her data set required around 47m30s/Msamples when BEAGLE was used on CPU (8cores).

 

It was actually somewhat faster without BEAGLE: between 30-37m/Msamples on 8 cores.

 

The flaw was entirely mine, and my apologies go to Marianne and everyone else for the lost time.

 

Mark