Computing nodes
Computing nodes can only be accessed with the job manager
All nodes have 4Gb of memory per core
to know the state of all nodes, use the check-cluster
command
Ciclad cluster (2021/01/20)
- 20 Computing nodes 1152 cores
- 4 nodes with 32 cores 128Gb AMD Opteron
- 16 nodes with 64 cores 256Gb AMD Opteron
Submit limit for Users is 1000 jobs
Standard User limit for running jobs : if any parameter is over limit jobs are waiting to be under limit to launch the jobs MAXNODE=2 MAXPROC=66 MAXJOBS=66 MAXMEM=252G
Parallel Jobs : Walltime MAX 1 Week ( 168H ) interactive jobs: Walltime MAX 6H
External Users ( Not in IPSL ) have one more limit
MAXPROC=172 for all external user
Climserv cluster (2021/01/20)
- 14 Computing nodes 896 cores
- 14 Computing nodes with 64 cores 256Gb AMD Opteron
Submit limit for Users is 1000 jobs
Standard User limit for running jobs : if any parameter is over limit jobs are waiting to be under limit to launch the jobs MAXNODE=4 MAXPROC=128 MAXJOBS=66 MAXMEM=504G
Parallel Jobs : Walltime MAX 1 Week ( 168H ) interactive jobs: Walltime MAX 6H
External Users ( Not in IPSL ) have one more global limit
MAXJOB=48 MAXNODE=3 MAXPROC=66
Hal GPU cluster
- 4 computing nodes with 2 gpu card
- 8 core Intel(R) Xeon(R) Silver 4112 CPU @ 2.60GHz
- 64Gb Memory
- 2 Nvidia GeForce RTX 2080 Ti