Configuring TIBCO® Data Science - Team Studio to use TIBCO Enterprise Runtime for R

TIBCO® Enterprise Runtime for R (TERR?) is a high-performance, enterprise-quality statistical engine compatible with the R language. To scale more effectively to larger data and higher performance, you can configure the R Server in your Spotfire® Data Science installation to use the TERR engine, instead of the open-source R engine, by following these steps.

Perform this task on the server where the installation of Team Studio is deployed.

1. Shut down the R-Execute server that is using R engines.

   cd /usr/alpine/r_connector  ./stop_services.sh

    
2. Install TERR? in the folder /usr/alpine/r_connector/TERR60GA.
3. Copy the following packages from /usr/lib64/R/library/ to /usr/alpine/r_connector/TERR60GA/site-library. (These packages are needed by R-Execute.)
   • data.table
   • Rserve
   • reshape2
   • chron
   • stringr
   • plyr
   • Rcpp

   NOTE:  Copy these installed libraries from an existing R installation, rather than installing them in TERR, because they must be compiled.

4. Make a copy of TERR.exe, place it in /usr/alpine/r_connector/TERR60GA/bin/R, and then rename it R.exe.
   
   This step is necessary because Rserve spawns engine processes by executing ".../bin/R CMD ..." (This behavior is hard-wired into the Rserve code with paste(file.path(R.home(),"bin","R"), "CMD", fn)).
5. Modify the R-execute startup script start_services.sh to create TERR_start_services.sh, substituting calls to R with calls to TERR.
6. Start the Team Studio R-execute service with TERR.

   cd /usr/alpine/r_connector  ./TERR_start_services.sh

    

Result

Using R-Execute (with TERR), you can run workflows made with R-Execute (with R), and generate the same results (captured text and output).

Notes

The data types integer, double, and string are handled correctly; however, logical and date values are input as strings.

It appears that R Server does not handle NA output well. If you see the error message, "REST upload from R server to Alpine failed, status code 500", then check whether your script contains code to output NA values.