Log for my R benchmarkme results

readme.md

benchmarkme Results

I periodically benchmark my computer systems in terms of their ability to carry out data analysis and related tasks in R. To do this I use the R benchmarkme package and I persist each set of results to a file (.rds to retain R class attribute) for posterity. I also log the run here at this very Gist.

I use the run_mark.R function, which I wrote based loosely on suggestions in the benchmarkme vignette, to do the benchmark tests and afterparty.

Machine description

Dell XPS 2011/Windows 10 (dual boot w/Debian)

• 2011 model Dell XPS L502x
• Intel i7-2720QM @ 2.20GHz, 8 cores
• 8 gb ram
• Windows 10
• Storage:
  • Samsung Evo 840 500 gb SSD (system drive)
  • Seagate 750 gb 7500 rpm (data drive in CD bay)

Dell XPS 2011/Debian Stretch (dual boot w/Win10)

Same specs as above but:

• Debian Stretch (testing), always with current package updates from the Debian repo
• RStudio updated from RStudio website, installed with Gdebi
• Storage: Seagate 750 only

Benchmark log

2016-10-11 Dell XPS 2011/Windows 10

• Initial benchmark run for this machine
• I am pleasantly surprised at this fairly old machine's middling rankings relative to other benchmarked machines
• I notice the benchmarking routine only accessed one core out of eight available on this machine
• R version 3.3.1 (2016-06-21) through RStudio 0.99.903
• BLAS optimize = FALSE

Test	My rank	N field
prog	114	173
matrix_cal	114	173
matrix_fun	71	173

2016-11-19 Dell XPS 2011/Windows 10

• Second benchmark for this machine. Performed very close to the initial run of a month ago.
• I need to start putting down the absolute results, not relative!
• R version 3.3.2 (2016-10-31) through RStudio 1.0.44
• BLAS optimize = FALSE

Test	My rank	N field
prog	112	173
matrix_cal	118	173
matrix_fun	80	173

2016-11-19 AWS EC2 running a t2.micro instance

• This is my personal Shiny/Rstudio Server cloud machine
• It's nearly the base level AWS cloud server
• It feels much faster in day-to-day use situations than my Dell XPS laptop, despite having 1/8 the memory and 1/8 the cores as my Dell machine
• The benchmarks are interesting! I didn't expect that this server would rank as highly as it did, and beat the pants off my machine in one of the tests
• Machine specs:
  • Ubuntu Server 16.04.1 LTS on GNU/Linux 4.4.0-47-generic x86_64
  • 1 gb ram
  • Intel(R) Xeon(R) CPU E5-2676 v3 @ 2.40GHz, 1 core, 64 bit
  • R version 3.3.2 (2016-10-31) through RStudio Server 1.0.44
  • BLAS optimize = FALSE

Test	My rank	N field
prog	72	173
matrix_cal	124	173
matrix_fun	116	173

run_mark.R

# Function to run the benchmark tests and spit out results to paste
#   into my log file

run_mark <- function(save_res = TRUE) {
	# Load the benchmarkme package
    if (require("benchmarkme")) {
        # Run the benchmark itself:
        res = benchmark_std(runs = 3)
        
        # Save the benchmark results out to an rds file
        if (save_res) {
        	the_path <- paste0('./benchmark_res_', Sys.Date(), '.rds')
            saveRDS(res, file = the_path)
        }
        
        # Do the results rankings and data sharing
        rank_results(res)
        upload_results(res)
        
        # Confirm pertinent info for my own log
        out <- list(r_version = get_r_version(), 
                    cpu = get_cpu(),
                    blas_opt = benchmarkmeData::is_blas_optimize(res),
                    ranks = plot(res))
        return(out)
    }
}
benchmark_results <- run_mark()
print(benchmark_results)