[Kappa-users] kappa: Service stop Tue Mar 30, 0800 CEST

[Per Lundqvist]

Kappa is now available again. Please note that the Kappa pilot testing
period now is over; Kappa is now in full production.

[Batch queuing system]

   The scheduler on Kappa uses fairshare (as Neolith does), but with a
   different scheduler and somewhat different underlying mechanism
   (for the interested:
   https://computing.llnl.gov/linux/slurm/priority_multifactor.html#fairshare).
   Current fairshare priority is visible with the command ``sprio''.
   
   Kappa compute time is divided equally between SNAC projects and
   Local LiU projects. In the future Kappa will be grid enabled and
   those jobs will take 3 % of the compute time on Kappa (while the
   remaining time will remain divided equally between SNAC and LiU).

   You currently have compute time on either one or more LiU projects
   or on one or more SNAC projects. Most of you have compute time on
   both a LiU project and on a SNAC project. Until now you haven't had
   to specify an account when running jobs on Kappa. But now all who
   are a member of different projects will have to, or else jobs will
   silently run as what's set as the default account.
   
   To list your configured accounts on Kappa use the command:
   
      my_accounts
   
   To specify an account when starting jobs add the argument:
   
      -A <account>
   
   to sbatch, srun, interactive, where <account> is the name of an
   account.

[Fat nodes]

   Kappa has 55 fat nodes (72 GB RAM) and 283 thin nodes (24 GB
   RAM). To use the fat nodes add the argument:
   
      -C fat
   
   to sbatch, srun, interactive.

[MPI]

   Kappa doesn't have Scali MPI, use either Open MPI or Intel MPI
   (available as the modules openmpi and impi respectively). Intel MPI
   might give higher performance results than Open MPI but isn't as
   well tested on Kappa.

regards,
/Kappa admin

-- 
Per Lundqvist

National Supercomputer Centre
Linköping University, Sweden

http://www.nsc.liu.se