[R] R 2.0.0 (Windows): slow startup over the network

[Prof Brian Ripley]

On Wed, 6 Oct 2004, Duncan Murdoch wrote:

> On Wed, 06 Oct 2004 15:50:34 +0200, Uwe Ligges
> <ligges at statistik.uni-dortmund.de> wrote :
> 
> >R-2.0.0 with the complete CRAN collection, Bioconductor and some other 
> >stuff installed on a capable Windows 2003 Server using a client with 
> >WinNT4.0:
> >The RGui window appears at once, the prompt within 5 seconds.
> >
> >
> >As Brian Ripley already pointed out, there must be a bottleneck at your 
> >side, either the server performance or network traffic.
> >I know from own experiences that WinNT 4.0 Server were not as good as 
> >the 2003 Server in file delivery, but I don't think it can be that dramatic.
> 
> I know almost nothing about Windows network administration.  Is there
> a way to log activity from a particular client, so you can see what
> it's requesting, and how long it takes to service the requests?

Given that we run a Samba server on Solaris (and on Linux) and only use a 
Windows 2000/2003 server for authentication and profiles, I did this 
from the client.  FileMon is one of a great set of utilities from
www.sysinternals.com which reports all file accesses, time-stamped.

Since we now have established that Thomas meant the default configuration
(not just base), 4 secs is tolerable and it looks like a server problem
(rather than a network problem). I would defragment the volume in case the
R\library directory has got badly fragmented.

For the record, R 2.0.0 took 2s to start on a Linux 500MHz Celeron box 
over our network, and takes 0.62s to start on a 2.6GHz P4 under Windows XP 
with a fast local disc (and about 0.38s on the same box under FC2 Linux).

-- 
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