Good computating practice
Ulf Ryde, Teoretisk kemi
Jan, 2004
This is a set of advices for the members of our group. The aim is to
facilitate
my job and to make your results available for other members of the
group
also after you have left the group.
- The jobs should be put in appropriate directories (one for each
project),
with a clear logic, and a clear name.
I recommend you to either start
the directories with a capital or use only capital letters for the
directories
and lower-case letters only for the files.
Never use only numbers for the
directories or for compounds - this is uncomprehensible for all others,
including your self some years in the future.
Quite often you do several (test) calculations on the same system,
leading to many directories. It is then very important that you mark
out which directory contains the final results (that go into the
publication). This directory should be the first directory you find in
the hierarchy. If you want to keep the history of the files, you can
either make a link in the parent directory directly to the final
results, e.g. using a directory Final or Best, or by making a file
giving the position of the best results.
Directories with wrong or temporary results should either be deleted or
marked Wrong or Bad or X_wrong.
Correct, but not optimal results should be marked Worse or X-worse.
This must be done when doing the calculations - otherwise you will
forget it and never do it. Thus, as soon as you get new results of the
same system, you should directly rename the directories so that it is
clear which contains the better of preferred results.
- All jobs should have a correct title (I realise that this is not
possible for large collections of calculations - then the directory
names are more important), indicating (for QM calculations)
- The system
- Charge
- Multiplicity
- Symmetry, if any
- Method
- Basis set
- Other interesting information
- Date, including year
- Start each working day by going through all your running jobs to
see if
they are finished or crashed. Check all running jobs at least once each
week that they still give sensible results.
- When a job has finished:
- Move the results to the back-uped file server (signe or garm).
- Run a Mulliken analysis (mul or mul2)
- Run describe to describe the geometry around the metals
- Make a table (one for each project, one line for each
calculation) with
job name, energy, geometry around the metals, and other interesting
information
(se we can easily see what jobs have been run). Use a StarOffice spread
sheet.
- Remove unnecessary files (purtur1)
- Gzip the rest of the files
- When the project is finished and published, move the files to a
CD and
to a /temp disk.
- Make a complete backup of all your files once each week (e.g. on
Friday afternoon, when you leave).
You will get a external hard disk for that.
- Always when you copy files, use cp -p to preserve the date of the
files.
- Save all your article pdf files in the following format:
journal-vol-year-page-author-keyword1-keyword2.pdf
e.g. jctc-5-09-396-mulholland-qmmm-dna.pdf
Note that only two digits for year.
Regularly copy all your files to /away/bio/Articles
There, you also find articles from other people in our group
- Seminars should be copied to /away/bio/Presentations