Louhi is intended for jobs that parallelize well. The standard procedure described below allows a user to run jobs of size 32-2048 cores, according to user specific limits based on scalability tests. For projects designated as Grand Challenge tasks and in special cases, the restrictions for the job size or running time can be altered.

If you are a new customer for CSC's computing and application software services, see first instructions for new customers below. The application procedure depends on whether your are an academic user or not.

For new CSC customers (in English)
Tervetuloa CSC:n asiakkaaksi (in Finnish)

Academic customers can apply for Louhi user accounts according to instructions on sites:

How to apply for academic user accounts
CSC:n tietokoneiden käyttölupien hakeminen

If you already have a user account on some other CSC's computing server (Corona, Murska), you will still have to provide scalabity test results for the application you will be running on Louhi. Contact usermgr at csc.fi to obtain a user account for Louhi to be able to run the scalability test. See instructions on how to run the tests and submit the results at:

Instructions for scalability tests on Louhi
Ohjeet skaalautuvuustestien suorittamisesta Louhella

Scalability test reports are sent back to parallel computing specialists (rinnakkaislaskenta at csc.fi).

All usage on Louhi will fall under the category of large applications. Initial CPU time resources (joint quota for all CSC's computing servers) are requested when you open your computing project. Starting quota for new projects is 5000 billing units. Additional CPU time is applied by the web form at:

Applying for CPU time (in English)
Laskentaresurssien hakeminen (in Finnish)

Computing resource applications are handled by the CSC Resource Allocation Group.

Currently, the configuration of queuing system is in progress and it is first tested in order to reach a stable and load-balanced configuration. After that it will be informed what are the maximum running times in different queues and for different number of cores.

Currently, the maximum running time for jobs targeted for 32-128 or 129-256 cores is 72 hours, and for jobs targeted for 257-512 or 513-2048 cores 36 hours. The default running time is four hours. The default memory is 1 GB/PE. A PE (Processing Element) is an intance of ALPS launched excutable such as a MPI task of an MPI job or a single instance of a sequential executable. Normally one PE is running in one core. The maximum memory request per PE is 2 GB, but it can be bypassed (see Chapter Parallel batch jobs). For testing purposes there is a queue for 1-128 cores (maximum running time 15 minutes) and for debugging 8 interactive XT4 nodes (32 cores) and 4 interactive XT5 nodes (32 cores).

The particular queue arrangements may change as deemed effective. See Section Monitoring and displaying job status for details on how to find the current queue configuration.