Parallel computation in R usable on windows machines with SOCKS parallelization.

parallel_example_in_r.r

library(parallel)

# detect number of compute cores
num_cores <- detectCores()

# register the cluster
cl <- makeCluster(num_cores)

# load libraries on all nodes in cluster
clusterEvalQ(cl, library(data.table))

# register pre-calculated (global) variable on all nodes in cluster

shared_dt <- data.table(a=1:10000, b=1:10000) 
clusterExport(cl, "shared_dt")

# split the indeces to delegate work to the nodes
indeces <- splitIndices(nrow(shared_dt), num_cores)

# parallel computation
result_list <- parLapply(cl, indeces, function(inds){
  # do the work for the selected part of data table
  res <- shared_dt[inds, c := a + b][inds,]
})

# row bind the results
result <- rbindlist(result_list)
result

stopCluster(cl)