[GRASSLIST:361] Re: GRID - GRASS - R

[Roger Bivand]

On Wed, 11 Jun 2003, Bernino Lind wrote:

> Hello GRASS users,
> 
> Does anyone know if there exist a project or lobbyism to make GRASS with
> (without) R run on the forthcoming GRID from CERN?
> 
> links:
> http://eu-datagrid.web.cern.ch/eu-datagrid/
> http://www.gridforum.org/

Some few attempts have been made to run R and GRASS on networks of 
computers. It is not however clear that - given rapid increases in 
single-node compute power over the last ten years - the burden of moving 
code to manually or automatically tile jobs into chunks for different 
processors, or to insert threading into pre-threading codebases, is likely 
to be undertaken. The more user interaction needed, the less sense there 
is in using many processors. In statistics, there are further issues, like 
ensuring that random number generators do not run synchronsed on different 
platforms. Probably GRID will be more useful for modelling (simulation 
and deterministic), rendering, and similar compute-intensive tasks, so 
links are much more likely to be loose-coupled than to involve migrating R 
or GRASS to such a compute environment. Databases, webservers, etc. are 
another matter, and if GIS becomes middleware for databases, then this 
assessment may be too negative.

Roger Bivand

Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Breiviksveien 40, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 93 93
e-mail: Roger.Bivand at nhh.no