ClusterComputingWithR.pdf

1Stowers Institute for Medical Research R/Bioconductor Discussion Group Earl F. Glynn Scientific Programmer 18 December 2007 Cluster Computing With R 2Cluster Computing With R ? Accessing Linux Boxes from Windows ? Linux Servers ? Cluster Head Nodes ? Linux Command Prompt ? xterms ? Linux Home Directories ? Setup For Examples ? R Batch Jobs Using Linux ? Cluster Basics ? Sun Grid Engine ? R Batch Jobs on Cluster: Toy Examples and Time Perturbation Study 3Accessing Linux Boxes from Windows Wiki: http://research/PuTTYxming PuTTY / Xming will be used to access Linux boxes from Windows 4PuTTY / xming Accessing Linux Hosts Right Click Xming icon genekc02 64-bit genekc03 32-bit Can access via PuTTY icon Accessing Linux Boxes from Windows 5Accessing Cluster Head Nodes Right Click Xming icon Confusing Names 32-bit cluster nodes cluster01 = Betelgeuse cluster02 = Sirius 64-bit cluster nodes cluster03 = Orion PuTTY / xming Accessing Linux Boxes from Windows 6/applications/astronomer/Atlas/NovDecJan.html 7Linux Console and Command Prompt cat .bashrc # or in .profile case $TERM in xterm*) PROMPT_COMMAND=echo -ne \033]0;${USER} ${HOSTNAME}\007 ;; *) ;; esac BLUETEXT=\[\e[34;1m\] RESETTEXT=\[\e[0m\] PS1=$BLUETEXT\n[\! \$(date +%d%b%y %T) \$( pwd )]\n$RESETTEXT Making Your Linux Bash Command Prompt Useful /tutorials/Matt/CmdPrompt.html Accessing Linux Boxes from Windows 8Linux Command Prompt and xterms Accessing Linux Boxes from Windows 9LINUX Home Directories Accessing Linux Boxes from Windows ? Files for cluster jobs normally should be in your home directory, e.g., /home/efg ? Suggest single directory per cluster job, e.g., /home/efg/cluster/R/simplegraph ? LINUX home directories can be accessed via Windows using an UNC name: \\lnaskc01\unixhomes\efg\cluster\R\simplegraph ? Not all filesystems available on genekc02, ..., are available on clusters, e.g., $BLASTDB 10 Summary Accessing Linux Boxes from Windows ? Use Linux boxes for R batch jobs (discussed next) ? Normall