Your accounts on bayes have been created. By default, they use
your nsid and password on paws for bayes. You can change your
password with terminal command "passwd" if you want
after you use ssh client (Putty
on windows, or terminal on mac) to log into bayes.usask.ca.
Use VPN to access bayes from off-campus. As required by the U of S and CFI, you can only use bayes on campus. If you need to use bayes from home, you need to download vpn client (http://www.usask.ca/ict/services/network-services/vpn/index.php). It may be very slow though to use bayes from home.
To use R, you can use rstudio server with your browser, such as firefox, safari. The url to log into rstudio on bayes is:
rstudio is only for developing and testing your R code. It cannot do parallel computing.
To understand basic linux terminal (shell) commands, you are
recommended to read this webpage:
You are in terminal if the line starts with $, which is different from R console that starts a line with > unless you configure it to be something else (if you know how to do so, you do not need to read this file).
log in bayes using ssh client (putty on windows, terminal in linux and mac) with your user and password. To transfer files between bayes and your desktop, you need to install "winscp" on windows, or "macfusion" on macOS. Rstudio web interface provides the function to transfer files too, but that is less efficient than winscp or macfusion.
suppose you want to run test.r simultaneously 100 times, do this:
upload the “test.r” file (http://math.usask.ca/~longhai/useBayes/bayesfiles/test.r) to a directory, for example ``test'' under your home directory ((which you need to create with mkdir test)
change to test directory from your home directory ~ in terminal($), type: cd test
Run test.r 100 times:
cvqsubR test.r 100
Explanation of cvqsubR test.r 100: It creates 100 R jobs by adding an R expression “ifold <- i” (for i = 1, …, 100) to the top of “test.r” then submit the 100 modified “test.r” to the cluster. All the 100 new “test.r” will be run simultaneously. To distinguish the 100 job outputs, use “ifold” as the identity in each job. More explanations are contained in “test.r” file.
qstat: to monitor status of jobs
qdel 100: to kill the job with id 100
mqdel: a shortcut to kill multiple jobs, for example
mqdel 300 100
The above command will kill jobs from id 300 back to 100 jobs, that is jobs 300, 299, ..., 201. Find the ids of your jobs using qstat.
for more introductions, you can visit the help page for cluster plato which uses the same scheduling system: http://www.usask.ca/ict/services/research-technologies/advanced-computing/plato/running-jobs.php. note that: You don't know need to understand the above description to use parallel computing. The fcommand "cvqsubR" creates a shell file based on your R code and submits it to bayes scheduler.
Type which cvqsubR, then read the file for details. It has two more arguments to specify the memory and walltime needed for your jobs. Using less walltime and less memory than the default settings may place your jobs in higher priority.