I use package multicore, and it works very well. There is, however, one thing I wonder if I don't do correctly, here is one example: I read ca 5.5 mill records into a dataframe, using read.dta through a one-line function rdam1. It runs nicely in parallel with other activitites. p1 <- parallel(rdam1()) ; Then I recover the data frame with collect: am1 <- (collect(p1))[[1]] All goes well, but the collecting step takes quite some time, and as the data frame is already in memory, should this be necessary? I experience the same with all parallelized steps resulting in large objects. Is there a way to avoid this, for instance, accessing the object through a pointer? Trond Ydersbond [[alternative HTML version deleted]]