[GRASSLIST:6146] Re: Information research lab

[Glynn Clements]

Ziad Shaaban wrote:

> I am preparing a proposal for a lab donation where the basic idea is the
> have a Beowulf clustering (using around 27 machines) to be used for GIS
> image processing applications.
> After a small research I found out that Grass is the best to fit.
> Have anyone done something similar and all help and references are welcomed.

One point to bear in mind is that it isn't particularly easy to get
GRASS to take advantage of multiple CPUs (e.g. SMP or clusters).

Apart from the fact that GRASS doesn't use multiple threads, running
multiple concurrent processes on a single dataset is complicated by
the fact that no locking is performed on any of the data files.

-- 
Glynn Clements <glynn at gclements.plus.com>