High-performance computing in R at useR! 2008

The useR! 2008 meeting is about to commence. Although I am not able to go this year I will be keeping a close eye on the talks and slides that (I assume) will be posted.  Last years useR! meeting (which I attended) was a great experience and considering the list of participants for this years meeting, it is bound to be a very interesting meeting. Dirk Eddelbuettel has already posted the slides to his talk entitled “Introduction to High-Performance R” on his web site. Clearly, anyone using R on, say, computer clusters would find this highly relevant.

On a side note, it would be interesting to have some idea of how commonly R is used on computer clusters. I have been using R on our in-house cluster for a few years now, and although it may sound a bit counterintuitive to run a high-level interpreted language such as R in a high-performance computing environment such as a cluster (most people would use a low-level compiled language like C for example), using R has the big advantage of allowing very fast prototyping. In other words, although the R code runs slower on the cluster than native C code would do it takes much less time to code it. Of course, for certain classes of models there is no other option that using C (or a similar low-level language), but if you can get away with R, a cluster allows you to run your serial code (or parallelized R code) in parallel with potentially huge time savings.

< size="2">This is from the “Mario’s Entangled Bank” blog ( < >< size="2">http://pineda-krch.com< >< size="2"> ) of Mario Pineda-Krch, a theoretical biologist at the University of California, Davis.< >