[R-sig-Geo] Revolution R Enterprise - solution to handling large (spatial) data?

[Roger Bivand]

On Fri, 10 Sep 2010, Tomislav Hengl wrote:

>
> (Maybe this is not really suited for the R-sig-geo, but since we often 
> experience problems with loading and visualizing large data)
>
> Has anyone yet used the Revolution R Entrerprise Version of R? Apparently it 
> fixes/reduces the problem of large data and has an excellent GUI.

No, it provides 64-bit support, which R for all platforms already does - 
the user has to provide the RAM as before. It does ship with a more 
integrated support for using multiple cores, but those facilities are also 
available for R anyway, provided the relevant packages are installed and 
loaded.

If you are thinking of "NEW! Big Data Analysis for Terabyte-Class File 
Structures — A comprehensive solution that provides fast, scalable 
statistical analysis of large data sets without the RAM barrier of 
standard R" - this does not apply to the 2G limit on single R objects, 
which for most geodata problems are addressed quite adequately by tiling, 
as in the raster package.

What it provides apparently is a visual IDE and debugging, in addition to 
support. It does not provide contributed packages beyond those Revolution 
Analytics themselves write and maintain (they do, they are pro-active 
people, and are doing a good job in relation to business).

If your university will not let you install regular R, this is a good 
solution, because it gives IT departments a "supplier". Certainly worth 
trying for comparison, but on the same hardware it is unlikely to perform 
differently from R itself (since they cannot fork R and leave GPL, so any 
dramatic breakthroughs in performance would be merged back into trunk, if 
they work cross-platform). The RevoScaleR package is not on CRAN, and 
simply provides facilities similar to those such as biglm, ff and sqldf 
for doing analyses of chunks of large data frames using a different XDF 
file structure. Present status is summarised in the Large memory and 
out-of-memory data section of the HPC task view.

Hope this helps,

Roger

>
> [http://www.revolutionanalytics.com/downloads/free-academic.php]
>
> Read more:
>
> "Startup wants to be R alternative to IBM, SAS"
> [http://www.networkworld.com/news/2010/050410-startup-wants-to-be-r.html?hpg1=bn]
>
> "A Community Site for R – Sponsored by Revolution Analytics"
> [http://www.inside-r.org]
>
>
> T. Hengl
> http://spatial-analyst.net
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>

-- 
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no