This is a guide aimed at you as a new user at our clusters. As a new user you are obliged to attend our introduction seminars that we hold about several times per semester. You can view information about the available clusters Alvis and Vera.
The cluster environment¶
When connecting to a computer cluster the actual computer you have connected to is only part of this system - the login node of the cluster. This is the computer in the cluster with which you mostly will work directly with. But it is not the computer that is doing the heavy calculations! You share this computer with a hundred other researchers. Also connected to the cluster is a large amount of "compute nodes" that do the actual big computations.
Access to the compute nodes are done via with a "batch processing system". The basic idea is that all computation tasks are packed and specified in advance, and then submitted as a "job" to the job scheduler of the batch processing system. The job scheduler will then put the job in its "job queue" and when the requested of computer resources are available it will execute the job on a partition of all the compute nodes of the cluster.
The result of this approach is a more efficient use of all the compute nodes of the cluster as jobs are scheduled and executed automatically without any required user intervention. Thus the cluster is working day and night, weekends and holidays as long as there are jobs in the queue.
The clusters use SLURM. There are numerous commands you use to submit and check your jobs.
Step by step¶
All new users need to:
- First you must get access to one of our resources.
- Connect to the cluster via SSH, Thinlinc, or the portal.
- Instructions for the filesystem and data transfer.
- How to access the software
- Running jobs
If you encounter problem, please see the the support page