Unexpected perl-run success

One of the responses to my question about fast computers was an offer to use a computer cluster at Dalhousie University. The guys in charge (Rob and Karl) set me up with an account and a password but I haven't had time to try it out until this afternoon.

Rob had tried to run our program and run into a problem - it wouldn't read the desired settings from the settings input file. This appeared to be due to some idiosyncracy with how our code was being interpreted by their perl compiler. When he tweaked the code to use a slightly different command that did the same thing, it worked. But I wasn't looking forward to having to go through our code and make these changes.

Karl had thoughtfully sent me some instructions when he sent me my password:

The cluster uses the `module` system for loading and unloading the environment variables required for the different applications, quick how to is as follows

Show the currently installed modules
`module avail`

Load the module for this shell only, not future shells
`module load mauve/2.2.0`

Load the module for all future shells
`module initadd mauve/2.2.0`

Remove the module
`module initrm mauve/2.2.0`

Beyond that pretty standard SGE install.

Unfortunately this was gibberish to me, as I don't know what a 'module' or a 'shell' is, or how to do a standard SGE install. I also couldn't understand the online documentation for remote use, which talked about things I'd never heard of, such as RSA auth KeyFiles and NX CLIENTs and Gnome.

But wotthehell, I figured I might as well try doing the same things I do to install and run my program on our local WestGrid cluster. So I used Fugu to log on to Rob and Karl's cluster (that worked!) and to copy my program and its two input files into my cluster directory (that worked too!). Then I opened a Mac Terminal window, used ssh to log on to the cluster (that worked!) and told it to run my program. Based on Rob's experience I was expecting a string of error messages, but the program ran fine! In fact it ran about 60% faster than it does on my laptop, which is about five times faster than it runs on WestGrid.

Now I just have to decide what runs need to be done. This will require sorting through everything I did while I was away and what I learned from it, and then consulting with my former-post-doc co-author. Expect a post tomorrow describing this.