[R] Running R from CD?

[Prof Brian Ripley]

On Mon, 22 Nov 2004, Prof Brian Ripley wrote:

[...]

> BTW, I believe running R 2.0.x from a CD will be a lot slower than 1.9.1
> because of lazy loading and frequent file accesses: that's a theoretical 
> issue we intend to address for 2.1.0, but not one anyone has yet commented 
> that it is a problem.

I collected some data (under Windows XP).

On a modern desktop, running R from a CD-R or from a USB 2.0 thumbdrive 
was perfectably acceptable, with startup times of about 5 secs and little 
delay when running.

On a 2.5year old laptop with a USB 1.1 port (but the same thumbdrive) it 
took about 15secs to start and with frequent delays the first time an 
object was used -- I would not find that tolerable.  The laptop's CD drive 
was slower than the desktop and there were delays when it powered down, 
but it was acceptable.

This was less performance penalty than I was expecting, and less than I 
have seen on a high-latency network file system. So it looks as if all we 
can do is trade a slower startup time (by caching files) for removing 
hiatuses when running.  (Caching the pkg.rdb and pkg.rdx files when a 
package is opened would probably only take up a little over 1Mb in a 
typical session.)

Writing to the thumbdrive took about 20mins, as R has so many small files
and the drive has a VFAT file system.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595