Skip to content

Cluster and parallelization utilities that came ouf of the Portage SMT project — Outils de parallélisation sur grappe de calcul issus du projet Portage de TAS

License

Notifications You must be signed in to change notification settings

nrc-cnrc/PortageClusterUtils

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Français

Portage Cluster Utilities

This repo contains scripts that are used to facilitate parallelization of jobs on a cluster or on a multi-core machine. These scripts were originally written as part of the Portage Statistical Machine Translation, but were extracted here because they are of more general interest.

Installation

The simplest installation procedure is to activate PortageClusterUtils in place from a clone.

git clone https://github.com/nrc-cnrc/PortageClusterUtils.git

and then add this line to your .profile or .bashrc:

source /path/to/PortageClusterUtils/SETUP.bash

Alternatively, you can also install it to the destination of your choice like this:

cd bin/
make install INSTALL_DIR=/install/path

which will copy all the scripts into /install/path/bin/. By default, the destination in $HOME/bin.

Dependencies

PortageClusterUtilities requires:

  • Perl >= 5.14, as perl on your PATH;
  • any version of Python 3, as python3 on your PATH;

Usage

Main Scripts

The main tools provided by this repository are these scripts:

  • parallelize.pl: take any non-parallel pipeline processing tool, say, a tokenizer, and parallelize it. E.g., parallelize.pl -n 10 'utokenize.pl < input > output' will produce the same results as utokenize.pl < input > output but it will run it 10-ways parallel, either on 10 cores, or in 10 scheduled cluster jobs if you're running on a cluster.

    Run "parallelize.pl -h" for more details.

  • run-parallel.sh: given a list of M bash commands to run, independently from one another, launch N worker jobs (scheduled jobs on a cluster or worker threads on a multi-core machine) which will run the all the jobs, N-ways parallel, until all are done.

    Run "run-parallel.sh -h" for more details.

  • psub: different computing clusters require different syntax to submit jobs. Those differences are encapsulated inside psub, so that run-parallel.sh does not have to be aware of them. psub -mem 24G -cpus 4 some-command -and -its -options will run "some-command -and -its -options" in a job with the requested resources.

    Currently, psub is highly specicialized the the clusters we use at the NRC. To use PortageClusterUtils on your cluster, you must adapt psub to write job scripts compatible with your cluster configuration.

    Run "psub -h" for more details.

Other Scripts

Script Brief Description
analyze-run-parallel-log.pl Summarize started/done/failed jobs in a run-parallel.sh log.
jobsig.pl Send a signal to a job.
jobtree Display the jobstat output as a tree of jobs (jobsub).
on-cluster.sh Detect if we're running on a cluster.
process-memory-usage.pl Tally the memory usage of a process tree.
qstatdir Run qstat, showing where commands were run from (qsub).
qstatn Run qstat -n with a more compact output (qsub).
qstattree Display the qstat output as a tree of jobs (qsub).
r-parallel-d.pl Daemon for run-parallel.sh.
r-parallel-worker.pl Worker for run-parallel.sh.
r-scheduler.py Monitor run-parallel.sh and maximize cluster usage (obsolete)
rp-mon-totals.pl Helper for run-parallel.sh, for tallying run-time stats.
rsync-with-restart.sh For really unstable connections, rsync with retries until success. Warning: never gives up!
stripe.py Helper for parallelize.pl.
sum.pl Sum/avg/max a column or list of numbers.
which-test.sh "which" with reliable exit code, for scripting.

Each script accepts the -h option to output its full documentation.

Scripts with a cluster/scheduler type in parentheses might only work on such a cluster.

Citation

@misc{Portage_Cluster_Utils,
author = {Joanis, Eric and Stewart, Darlene and Larkin, Samuel and Leger, Serge},
license = {MIT},
title = {{Portage Cluster Utils}},
url = {https://github.com/nrc-cnrc/PortageClusterUtils}
year = {2022},
}

Copyright

Traitement multilingue de textes / Multilingual Text Processing
Centre de recherche en technologies numériques / Digital Technologies Research Centre
Conseil national de recherches Canada / National Research Council Canada
Copyright 2022, Sa Majesté le Roi du Chef du Canada / His Majesty the King in Right of Canada
Published under the MIT License (see LICENSE)

About

Cluster and parallelization utilities that came ouf of the Portage SMT project — Outils de parallélisation sur grappe de calcul issus du projet Portage de TAS

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •