I also used a very similar method for the Bruker instruments, with
the exception that a shell script ran in the background and checked
for acquisition processes ("go") at 2 minute intervals. If any were
detected, they were "timed" by seeing how many 2 minute intervals
were spanned. The output is a list of the user, the experiment run,
number of scans, number of increments (obtained from the acqus file),
and the elapsed time. A cron job runs periodically to see if the
script is still running, and starts it if it is not.
It is not horribly complicated to write, but does involve several C
programs in addition to the shell script - sometimes things had to
be adjusted for differing versions of the NMR software. I found it to
be pretty accurate, and a summary program (written in C) shows the
overall usage of the instrument over a given period of time in terms
of the various users. If anyone wants further details, please let me
know.
Tom Pratum
On Jul 1, 2005, at 1:56 PM, Steve Philson wrote:
> On Fri, Jul 01, 2005 at 08:13:35AM -0400, Hopson, Russell wrote:
>
>>
>> Similar to recent request for SGI accounting software, I am
>> looking for
>> Linux accounting software to track NMR spectrometer use.
>> Unfortunately,
>> the wtmp files that are generated each month do not accurately
>> account
>> for the time the users are on logged into the workstation, and I
>> am not
>> sure why.
>>
>> I would be very interested to hear how other facility managers track
>> users time for billing purposes (ie. Reservation systems, actual
>> workstation logs, etc).
>>
>
> For a long time I have used cron jobs that run in the middle of
> every 10 minute period (our scheduling and billing increment); if
> the ps output shows the process characteristic of operation of the
> spectrometer, the name of the owner is written to a file (otherwise
> just a blank line). Each day at midnight another cron job changes
> the file, so I have a set of files, labeled by the day (e.g., 050629)
> with 144 lines, each either blank or with the name of the user. (I
> write the day of the week to the first line, since our billing is
> different for weekends, so they are actually 145 lines long.)
>
> Back when I started this it was because on our solid state instrument
> (Chemagnetics, SunOS 4) people weren't scheduling time (my normal
> way of doing the billing), and they also were also using the nmr
> program for processing separately from running the spectrometer,
> and only the latter was to be charged.
>
> Later I began doing the same thing on our Varian instruments
> (Solaris). Even though there our setup allowed only one login at
> a time, the "last" command was very unreliable under early Solaris
> versions, and the more complicated process accounting seemed like
> way too much, and still difficult to use.
>
> I wrote some simple perl scripts to combine the scheduling and
> logging information and convert it to billing results. Since we
> charge differently according to time of day and day of week, it is
> fairly simple to assign a charge for each slot and just add them up.
> It is not perfectly accurate, but it is close enough for us.
>
> --
> Steve Philson philson_at_nmr.chem.umn.edu
> Director NMR Lab 612-626-0297
> Chemistry Dept. University of Minnesota
>
>
Tom Pratum
Dept of Chemistry, MS 9150
Western Washington University
pratum_at_chem.wwu.edu
Received on Wed Jul 06 2005 - 12:37:58 MST