First aff all, you can't just say "besides being free", it's not "free",
it's Free Software, which is not the same.
It's not about if it's technically superior, it's ethically Superior, whichs
is failrñy enough for me to use it (i would be using GNU/Linux even if it
were thecnically inferior, and once upon a time, GNU was a little starting
disaster, and i used it anyway)

BTW: about the thech part of the discussion, i just don't understand how can
you use windoze and say it's stable (even XP). And i knew what i'm talking
about, i work for a hosting company, and we have, besides our GNU/Linux
server, windows 2k servers, and XP machines at the office.


beliavsky@aol.com wrote:

I used Unix workstations in graduate school for scientific
programming, but since then I have used flavors of Windows. When I was
running Windows 95 or 98, the case for moving to Linux was obvious --
it's better to use an operating system that does not crash every few
hours (yet I muddled through with Windows).

I have found Windows XP to be very stable. I have good Fortran 95
compilers (Lahey, Compaq) and can use many tools that originated on
the Unix side, such as Python, Perl, Gnuplot, and Emacs. The Windows
command line interpreter is less powerful than Linux/Unix shells such
as bash or korn, but for any nontrivial script I can use Python on
either platform.

Besides being free, what compelling advantage(s) does Linux have over
Windows XP for number-crunching?

On Sun, 30 Nov 2003 16:51:37 -0800, beliavsky wrote:

Scalability. Distributing jobs across lot's of *nix machines (applies to
both Unix and Linux) is easy. You can also run on lots of machines
simultaneously, I routinely haves an Emacs shell open on several systems
at once, including systems on the other side of the continent.

Easy access from anywhere. With SSH you can be anywhere in the world and
open an Emacs or Xemacs window on your systems with very good performance.
Windows is so graphics heavy that running over a LAN, let alone a
broadband line, is painfull.

On 30 Nov 2003 16:51:37 -0800, beliavsky@aol.com wrote:

.... all the typical '...ilities': stability, scalability,
maintainability, securibility, interoperability;
no vendor lock in, runs on a wide range of hardware (platforms),
wealth of tools, open-ness, low-cost, technical support, etc
..
--
/// Michael J. Tobler: motorcyclist, surfer, skydiver, \\\
\\\ and author: "Inside Linux", "C++ HowTo", "C++ Unleashed" ///
New crypt. See /usr/news/crypt.

beliavsky@aol.com wrote:

I somewhere read (don't ask me where) that Windows effectively leaves 90% of
the CPU cycles for number crunching, whereas Linux leaves 95%. I think the
reason for that was that under Windows the GUI creates more "background
load". I am not sure if Linux was run in graphical mode for that test,in
which case this argument would be quite useless. Although, Linux at least
allows you to turn down the GUI...

On Sun, 30 Nov 2003 16:51:37 +0000, beliavsk wrote:

My experience has been that Maple runs better under linux than Windows. I
don't have quantitative data on that, only my impressions from working
with both. Also, Mupad is a nice linux alternative for that sort of work
as well.

I don't have any experience with linux fortran programming, so can't help
you there.

One thing occurs to me: eventually you will want to write up
your work, and TeX is the clear winner for scientific papers. I know it
exists on Windows, too, but the linux version is better than any I have
tried under windows. LyX makes writing it easier, too, and is far more
flexible than Scientific Word or the like.

--

David L. Johnson

__o | It doesn't get any easier, you just go faster. --Greg LeMond
_`\(,_ |
(_)/ (_) |

Harald Grossauer wrote:
seriously io-limited database population job. That is why the system
time is so high. Even so, there is plenty of CPU time available to run
two instances of setiathome. When I am not doing much, the setiathome
processes get over 98% cpu time (each) (in top, the %CPU will add up to
200% if it is completely CPU limited). So, when idle, running only the
two instances of setiathome (total of 89 processes in the proc table,
with X, GNOME and sawfish running), only about 2% of the total cpu time
seems spent in navel contemplation. Below, the list of processes are
chewing up 198.8% (if I added correctly) of the total CPU power, so I
can infer the waste is only 1.2%, much less than Windows (95, anyway) of
10% busy when nothing is going on for the benefit of the users. The
setiathome processes are essentially compute limited (they read about
300k bytes every 12 hours each, and write next to nothing).


6:05pm up 37 days, 16:38, 2 users, load average: 3.91, 3.96, 3.93
107 processes: 104 sleeping, 3 running, 0 zombie, 0 stopped
CPU0 states: 95.2% user, 4.6% system, 86.6% nice, 0.0% idle
CPU1 states: 92.7% user, 6.6% system, 83.1% nice, 0.7% idle
Mem: 513656K av, 506252K used, 7404K free, 0K shrd, 48720K buff
Swap: 1048568K av, 56744K used, 991824K free 299312K cached

PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME COMMAND
19 19 15800 15M 840 R N 86.0 3.0 344:18 /home/seti1/bin/setiathome
19 19 14484 13M 584 R N 83.4 2.7 583:16 /home/seti1/bin/setiathome
9 0 22800 13M 12240 S 11.9 2.6 722:22 db2agent (STOCK)

9 0 3852 3852 2260 S 3.6 0.7 7:50 /home/jdbeyer/bin/enter_vl
19 18 1248 1248 1048 S N 2.8 0.2 2:08 xosview
9 0 11636 5800 5556 D 2.3 1.1 3:26 db2pclnr

9 0 46216 28M 3528 S 2.1 5.6 80:48 /usr/bin/X11/X :0 -auth
11 4 12360 11M 6128 S N 1.7 2.3 10:04 /usr/bin/python 28100
19 18 3500 3420 2888 S N 1.2 0.6 1:42 cpumemusage_applet
9 0 5468 5436 4068 S 1.0 1.0 0:48 panel --sm-config-prefix
9 0 7592 1756 1548 D 1.0 0.3 5:37 db2loggr

19 18 1120 1116 872 R N 0.6 0.2 3:31 /usr/bin/top
9 0 0 0 0 SW 0.5 0.0 480:53 kscand
9 0 3668 3648 3024 S 0.5 0.7 0:48 modemlights_applet
9 0 0 0 0 SW 0.1 0.0 1:26 kupdated
19 18 3088 3072 1892 S N 0.1 0.5 0:34 xterm -e nice -18


db2agent, db2pclnr, and db2loggr are database server processes and
enter_vl is a database client.

When all is said and done, though, what difference does 8% make? If you
are 8% too slow, you will probably need a faster system no matter what,
and if you are 8% faster, you can put off the new machine by only a
couple of weeks.

--
.~. Jean-David Beyer Registered Linux User 85642.
/V\ Registered Machine 73926.
/( )\ Shrewsbury, New Jersey http://counter.li.org
^^-^^ 5:55pm up 37 days, 16:28, 2 users, load average: 4.03, 4.09, 3.95