Page tree
Skip to end of metadata
Go to start of metadata

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

Introduction

We have added SSDs, solid state disks, on a few compute nodes:

  • These are local fast disks that can speed up applications that preform a lot of intensive I/O operations,
  • hence such jobs should complete faster when using SSDs.
  • Jobs that do not perform intensive I/Os should not use the SSDs - these are a scarce shared resource.

  • Since these disks are local to the compute nodes: 
    • you cannot see the SSD from either login nodes,
    • your job will be able to use the SSD only while the job is running, hence
    • you need to prepare the files needed for a job prior to submitting the job, using /pool or /scratch, and
    • request the right amount of SSD disk space:

      • this is the maximum amount of disk space your job will need on the SSD at run-time, (similarly to the maximum of memory it will need).

    • Your job script will have to 
      • copy those files to the SSD before processing them,
      • adapted to use the SSD,
      • upon completion, copy the results from the SSD elsewhere (/pool or /scratch),
      • and delete what you wrote on the SSD.
    • If you exceed the amount of disk space, i.e., your quota, your job won't be able to write any longer to the SSD, 
      hence your job script should stop when encountering such error.

(warning) Since we have only a few of them, they should be used only if your application greatly benefit from using an SSD.

(lightbulb) The SSDs are currently only available for the uTSSD.q queue, a restricted queue. Contact Sylvain or Rebecca if you want to be authorized to use it.

HowTo

Since you can't access the SSDs from a login node, you must prepare the data the job will need somewhere else, like on /pool (or on /scratch) before submitting a job.

Like for memory, you need to guestimate how much SSD space your job will need. You will not be able to use more SSD space than you requested.

(lightbulb) Remember, your job will still be able to access the /home, /data, /pool, and /scratch disks, hence you don't have to copy everything on the SSD,

only the I/O intensive part of the analysis should use the SSD.

How to Prepare my Data

  1. Create a subdirectory in /pool (or /scratch) and move or copy the data you will need, for example (as user smart1)
    cd /pool/genomics/smart1
    mkdir -p great/project/wild-cat
    Now is have a directory for this case, and would copy the I/O intensive part of the required data set in it.

  2. While not required, you can pack these data in a compressed tar-ball
    cd /pool/genomics/smart1/great/project/wild-cat
    tar cfz ../wild-cat.tgz .
    The file /pool/genomics/smart1/great/project/wild-cat.tgz now holds you input data set, 
    being compressed it is likely to be smaller than the content of /pool/genomics/smart1/great/project/wild-cat,
    That directory can be deleted, unless you will need it later.

  3. Ancillary data and/or configuration files that are not causing intensive I/O can stay on a location under /pool (or /scratch)

How to Adjust my Job Script to Use the SSD

1  Copy the Data to the SSD

    • At the top of your job script, load the tools/ssd module and copy or extract your data set as follows:
example using recursive copy
module load tools/ssd
cp -pR /pool/genomics/smart1/great/project/wild-cat/* $SSD_DIR/.

or

example using compressed tar-ball
module load tools/ssd
cd $SSD_DIR
tar xf /pool/genomics/smart1/great/project/wild-cat.tgz

(lightbulb) The advantage of the compressed tar-ball is that the .tgz file is likely to be smaller than the content of the directory, hence less I/O transfer from the /pool disk, while un-compressing and writing  to the SSD is fast,

2 Adjust the Script or a Configuration File

  • You need to replace all references to /pool/genomics/smart1/great/project/wild-cat by $SSD_DIR,
  • this can be easily done at the shell script level, but not in a configuration file, i.e., for flags/options
      execute -o /pool/genomics/smart1/great/project/wild-cat/result.dat
    is replaced by
      execute -o $SSD_DIR/result.dat

  • Here is a simple trick to modify a configuration file:
    • Let's assume that your analysis uses a file wow.conf, where for instance the full path of some files must be listed, like:

      wow.conf
      # this is the configuration file of the fabulous WOW package
      input=/pool/genomics/smart1/great/project/wild-cat/wiskers.dat
      output=/pool/genomics/smart1/great/project/wild-cat/tail.dat
      paws=4
      eyes=2
    • Replace the wow.conf file by a wow.gen file as follows:

      wow.gen
      # this is the configuration file of the fabulous WOW package
      input=XXXX/wiskers.dat
      output=XXXX/wild-cat/tail.dat
      paws=4
      eyes=2
    • And create the wow.conf file from the wow.gen at run-time by adding the following in the job script:

      sed"s=XXXX=$SSD_DIR="wow.gen > wow.conf

      As long as XXXX is not used for anything else, this will replace every occurrence of XXXX by the value of the environment variable SSD_DIR.

3 Run the Analysis

  • With your data copied to the SSD and with your commands and configuration files adjusted to use the SSD, run your analysis.

4 Copy the Results from the SSD

  • At the end of the job script, you must add instruction to copy the results of your analysis back to /pool(or /scratch, or /data).

    If/when the results are easily identifiable, you can use the commands mv or tar,  and find, here are a few examples:


    1. Move the directory where all the results are stored and the log file, delete the rest.

      moving identifiable results, delete the rest
      # move results and log file back
      cd $SSD_DIR
      mv results /pool/genomics/smart1/great/project/wild-cat/.
      mv wow.log /pool/genomics/smart1/great/project/wild-cat/.
      #
      # delete the rest
      rm -rf *
    2. Move the directory where all the results are stored and the log file, delete the input (conservative approach, in case you missed something).

      moving identifiable results, delete known input sets
      # move results and log file back
      cd $SSD_DIR
      mv results /pool/genomics/smart1/great/project/wild-cat/.
      mv wow.log /pool/genomics/smart1/great/project/wild-cat/.
      #
      # delete input set and other stuff
      rm -rf input
      rm wow.gen wow.conf
    3. Move using the --update flag of mv (see man mv)

      moving using --update
      # move results using --update
      cd $SSD_DIR
      mv --update * /pool/genomics/smart1/great/project/wild-cat/.
      #
      # delete the rest
      rm -rf *

       Note, you can use mv --update on an explicit list (of files, directories, or file specification), not just * (everything), and you do not have to remove the rest,  but can only remove what you know you can safely remove (conservative approach).
       

    4. Find newer files and move them: the trick is to create a 'timestamp' file before starting the analysis.
       That file can be used later to find any newer file with the --newer= option of tar (see man tar):

      Using a timestamp file and tar --newer=
      # set the timestamp
      date > $SSD_DIR/started.txt
      # run the analysis
      ...
      # copy the new files in the subdir data/ to a compressed tar-ball
      cd $SSD_DIR
      tar --newer=$SSD_DIR/started.txt -cfz /pool/genomics/smart1/great/project/wild-cat-results.tgz data/
      # now remove it
      rm -rf data/
      # etc...
      # delete everything, unless
      rm -rf *

      See previous comments and what to tar and what to remove: once you've tar'd new stuff in data/,  remove data/, etc.

    5. Using the timestamp file and  the find command (see man find):

      Using find and a timestamp file
      # set the timestamp
      date > $SSD_DIR/started.txt
      # run the analysis
      ...
      # find the new files in the subdir data/
      cd $SSD_DIR
      find data/ -newer $SSD_DIR/started.txt -type f > /tmp/list
      # do the same on logs/, append to the list
      find logs/ -newer $SSD_DIR/started.txt -type f >> /tmp/list
      # etc...
      # now save what is in the list with one tar
      tar --files-from=/tmp/list -cfz /pool/genomics/smart1/great/project/wild-cat-results.tgz data/
      # now remove data/ and logs/
      rm -rf data/ logs/
      # etc...
      # delete everything, unless
      rm -rf *


      (lightbulb) BTW, the advantage of writing a .tgz file, rather than moving files is two fold, assuming your stuff is compressible:(tick) There are many more ways to accomplish this ....

      1. You write less in the .tgz file, so it should be done faster (reading and compressing should be fast, writing is the slow step)

      2. you need less disk space for your output (since it is compressed).

The drawback being that you need to know how to handle/view/deal with a .tgz file.

A Pseudo Example

Here is what a job script might look like:

Pseudo Example
#
#$ -N example
#$ -o example.log -cwd -j y
#$ -q uTxlM.rq
#$ -l ssduse=2560G,hm,mres=20G,h_data=20G,h_vmem=20G
#
# pseudo example using a fake package WOW, on the SSD
#
echo $JOB_NAME started `date` on $HOSTNAME in $QUEUE jobID=$JOB_ID
#
module load tools/ssd
module load special/wow
#
# create a wow config file from a generic version, to insert the SSD temp dir value
sed "s=XXXX=$SSD_DIR=" ~/wow/wild-cat.gen > ~/wow/wild-cat.conf
#
# cd to the SSD temp dir and copy the data set to it, using the existing .tgz file
cd $SSD_DIR
tar xf /pool/genomics/smart1/great/project/wild-cat.tgz
#
# create some sub dirs for output and logs
mkdir output
mkdir logs
#
# run the wow analysis (note how some files are not on the SSD)
wow --type=m --params=$HOME/wow/parameters.dat --config=$HOME/wow/wild-cat.conf -o $SSD_DIR/output -l $SSD_DIR/logs
#
# save the output and the logs in a tar compressed file
# (assumes wow did not change current working directory)
# otherwise insert: cd $SSD_DIR
tar -cfz /pool/genomics/smart1/great/project/wild-cat-results.tgz output/ logs/
#
# remove everything (in $SSD_DIR), or remove what you know you can (conservative option)
rm -rf *
#
echo $JOB_NAME done `date`



Last Updated  SGK


  • No labels