USER INFORMATION, PARTITIONS QUOTA & ACCOUNTING
Accounts are set for users associated to a project. Each project can have several users.
The accounts are closed 3 months after completion of the project. The accounts, project and data in the filesystems are removed at that time.
2. Available filesystems
|$HOME||home directory for user|
|$PROJECT||storage of project related source code, binaries, etc.|
|$DATA||storage location for large data|
3. Storage limits per project
|$PROJECT||10 TB → Can be increased upon demand|
|$DATA||100 TB → Exclusive for large projects can be increased upon demand|
There are no backups for the filesystems. The user must download the data as soon as it is available.
5. Scheduler & Prioritization
|Job prioritization||fairshare, waiting time in queues, job size|
6. Nodes exclusivity
Default exclusivity of compute nodes starts at 144 cores (4 compute nodes).
User can set exclusivity for the first 4 compute nodes.
|private||exclusive for sys admin or national urgency runs (high priority)|
|short||10 nodes (360 cores) Maxtime=72 hours|
|medium||30 nodes (1080 cores) Maxtime=48 hours|
|debug||58 nodes (2088 cores) Maxtime=24 hours|
CPU time is charged for wall-clock time, that is, for the time nodes are occupied.
CPU time is charged for complete nodes, regardless the number of cores used.
Monthly average CPU time set for each project.
Sliding usage window of 2 months: after two months the hours of the first month that have not been consumed will be accounted and deducted from the total available compute hours. The sliding window is moved every two months, so hours will be deducted on the 2nd, 4th and 6th, etc, months of the project.
CPU time transited to other months will have a slower priority than the corresponding monthly time.
CPU time not consumed at the end of the project will not be usable, unless there is a project extension.