At the moment, the R Project app (more specifically the "r" search command) doesn't make use of the map/reduce functionality.
But you can use Splunk search commands like search, stats, tstats, etc that does the map/reduct tricks to collect data and then send the data to your R script.
So the script is only executed on the search head in the context of a R sub-process which is spawned every time the search is executed.
That means it's still native R environment which uses memory to perform computations. The app uses temporary files to send the input data from Splunk to R and the results back from R to Splunk. I'm thinking about changing that so that the data is streamed instead of copied in a whole.
