[R] choosing systems

Douglas Bates bates at stat.wisc.edu
Mon Jan 11 20:47:09 CET 1999


>>>>> "GH" == gordon harrington <gordon.harrington at uni.edu> writes:

  GH> After many helpful suggestions from many including Brian Ripley,
  GH> Peter Dalgaard, Martin Maechler, Douglas Bates, and Tristan
  GH> Savatier--it is clearly time to upgrade my system from SPARC-10
  GH> Solaris 2.4. Rational choice of configuration should be based on
  GH> intended application and perhaps the fact it is my pension, not
  GH> my university, which now finances my research.

  GH> My most demanding applications are R, Xlispstat, S(ancient
  GH> vanilla 2.0) and my own ad hoc f77 (and ratfor) programs. I do
  GH> Monte Carlo multivariate models of quantitative multigenic
  GH> phenomena with significant genetic-environmental interactions
  GH> and compare the results both numerically and graphically with
  GH> data-banked empirical data. On my present system an optimized
  GH> fortran run takes about three days.

  GH> Suggestions keep pointing me toward an Intel box with a PC disc
  GH> partition for those kinds of occasional uses and with linux
  GH> or(perhaps and) Intel-Solaris for the primary usage. Other than
  GH> for the substantial disc storage needs to be ported, I have
  GH> little idea of optimum configuration.for these systems in terms
  GH> of memory or other hardware elements. Cost seems to dictate
  GH> IDE. Look and feel might give an edge to Solaris but linux
  GH> offers a wider base of user experience to tap.

  GH> I would appreciate any suggestions from those with experience
  GH> with comparable application needs.

I think you are right that the price/performance of Intel-based
systems running Unix or Linux is hard to beat these days.  We use both
Solaris/Intel and Linux/Intel.  I prefer Linux as it has more
applications available for it.  It certainly has momentum in its favor
these days and I feel that the idea of free or Open-Source software
has a lot going for it.  In contrast, Solaris/Intel is in a peculiar
position.  Sun would like people to be using it but they _really_ want
people to be using Solaris/SPARC so they can sell them hardware.  This
means that support decisions for Solaris/Intel could be influenced by
commercial considerations rather than technical superiority.

If you are willing to take a day trip from Waterloo to Madison, I
would be pleased to show you what we are doing with Linux here.  If
you want to bring your CPU with you we can probably get Linux
installed on it while you are here.

I haven't had any performance problems with IDE disk drives.  I just
compiled a 2.2.0-pre5 kernel on Friday evening (the day before
2.2.0-pre6 came out - naturally) and it recognizes my drive/controller
combination can do DMA.  I would put the money into memory before
worrying about SCSI versus IDE.  Besides, a 6 Gb IDE drive sells for
about (US)$100 these days so the investment is not too great.

I would buy a larger hard drive so you can partition it and run
multiple operating systems if you choose.  The only real disadvantage
to having a lot of disk space is deciding how you are going to back it
up.

As for distributions of Linux, I think that RedHat is the easiest
system to work with on your own and to get support for.  I personally
prefer Debian but it is probably not the best choice to try to install
for the first time.  Make sure you get RedHat 5.2.  I have seen it
advertised in computer stores here for $20.  (Interestingly enough, it
was advertised as a "Utility".  I can just imagine the typical
computer neophyte who installs this "utility" and ignores the little
warning about "You are about to re-format your hard drive.  This may
result in a loss of data".)
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._



More information about the R-help mailing list