...
To display a list of fairshare and usage for accounts, use sshare -la
. This command will display fairshare information across all account in the SLURM cluster.
Code Block | ||
---|---|---|
| ||
[george@umhpcgeorge@login01]$ sshare -la Account User RawShares NormShares RawUsage NormUsage EffectvUsage FairShare LevelFS GrpTRESMins TRESRunMins -------------------- ---------- ---------- ----------- ----------- ----------- ------------- ---------- ---------- ------------------------------ ------------------------------ free george 1 0.500000 12805068 0.872021 0.872891 0.250000 0.572809 cpu=0,mem=0,energy=0,node=0,b+ |
...
Partition | CPU | Memory | GPU | MaxPerNode | |||||
---|---|---|---|---|---|---|---|---|---|
cpu-opteron | 468.75 | 125 | 0 | 30000 | cpu-epyc | 375 | 1500 | N/A | 36000 |
gpucpu-epyc-k10genoa | 656.25625 | 3751202625 | N/A | 2100080000 | |||||
gpu-k40c | 700 | 400 | 11200 | 22400 | |||||
gpu-titan | 750 | 200 | 12000 | 24000 | |||||
gpu-v100s | 1437.5 | 500 | 46000 | 92000 | |||||
gpu-a100 | 4687.5 | 600 | 15000 | 1200000 | |||||
gpu-a100-mig | 4687.5 | 600 | gpu:a100_4g.40gb=83500 gpu:a100_3g.40gb=66500 | 1200000 |
Each core allocated for non-multithreaded jobs will be treated as 2 CPUs and no multiple multithreaded jobs should fall within the same core. All jobs will be billed based on the highest amount of resource type allocated. For examples:
- A non-multithreaded, 48 cores and 32 GB memory job running in
cpu-epyc
:- The billing value of each type of resources can be breakdown as follow:
- CPU = 96 (2 CPUs per core, 48 cores) * 375 (Billing value per CPU in cpu-epyc) = 36000
RawUsage
per minute - Memory = 32 (32 GB memory) * 150 (Billing value per GB memory) = 4800
RawUsage
per minute
- CPU = 96 (2 CPUs per core, 48 cores) * 375 (Billing value per CPU in cpu-epyc) = 36000
- Therefore, the job will be billed for 36000
RawUsage
per minute as 48 cores has the highestRawUsage
per minute compared to 32 GB memory.
cpu-opteron
:- The billing value of each type of resources can be breakdown as follow:CPU = 32 (32 CPUs) * 468.75 (Billing value per CPU in cpu-opteron) = 15000
RawUsage
per minute - Memory = 240 (240 GB memory) * 125 (Billing value per GB memory in cpu-opteron) = 30000
RawUsage
per minute Thus, the job will be billed for 30000
RawUsage
per minute as 240 GB memory has the highestRawUsage
per minute compared to 32 CPUs. - The billing value of each type of resources can be breakdown as follow:
gpu-v100s
:- The billing value can be breakdown as follow:
- CPU = 4 (2 CPUs per core, 2 cores) * 1437.5 (Billing value per CPU in gpu-v100s) = 5750
RawUsage
per minute - Memory = 64 (64 GB memory) * 500 (Billing value per GB memory in gpu-v100s) = 32000
RawUsage
per minute - GPU = 2 (2 GPUs) * 46000 (Billing value per GPU in gpu-v100s) = 92000
RawUsage
per minute
- CPU = 4 (2 CPUs per core, 2 cores) * 1437.5 (Billing value per CPU in gpu-v100s) = 5750
- Hence, the job will be billed for 92000
RawUsage
per minute as 2 v100s GPUs has the highestRawUsage
per minute among 2 CPUs and 64 GB memory.
...