Here are two tables illustrating the job information and resource usage in terms of each Slurm account. All of the jobs data is taken from the production partition, and the data is refreshed every hour. Only accounts currently has running or pending jobs are shown in the table. If you do not know which Slurm account your group belongs to, please use the showLimits (https://www.vanderbilt.edu/accre/documentation/accre-commands/#showlimits )command to check.
Regarding the details in the table, Fairshare is usually number of cores that the group purchased. Fairshare represents the lower limit of core usage that Slurm should allocate for the account’s usage. On the other hand, because the computation resources are shared between accounts, users are able to take advantages of idle resources of other groups; hence the practical usage limit for a group usually can be much higher than the fairshare. MaxCPU characterizes the maximum bursting limit of the CPU usage, it’s the upper limit that an account can use at anytime.
How many cores are available for use is continuously calculated and tracked by Slurm. Because the difference between MaxCPU and Fairshare is the other accounts’ contribution, when the cluster usage becomes tight, the cores can be allocated will trend towards the Fairshare end. On the other hand if there are a lot of idle cores the real usage for an account could ideally trend towards the MaxCPU value.