CPU-hours per user
Table
Memory per user
Average resident memory (RSS) of each user's processes, in GiB. Shared pages count once per process, so this overstates memory for programs that share a lot.
Table
Load against cores in use
Cores in use is CPU time actually burned. Load average also counts tasks waiting on disk or NFS, which is why it sits far above the cores in use here.
Table
Utilisation
Cores in use as a share of the … cores on gluon.
Table
Slurm CPUs allocated per user
Average CPUs held by running Slurm jobs, from squeue. Allocated is not the same as used: compare with the CPU-hours chart.
Table (with running and pending jobs)
When is gluon busy?
Average cores in use by hour of the week, over every sampled hour (…). Times are gluon's local time.
Table
Bootstrap datasets from before the sampler
The sampler only sees what happens while it runs. These two sources reach further back, but measure different things.
Hours logged in per user, per month
From the system login records (wtmp), since …. Hours with at least one SSH or console session open; overlapping sessions count once. Being logged in is not the same as computing.
Table (every user)
CPU-hours since the last boot
Snapshot of the kernel's cgroup counters on …, covering everything since gluon booted on …. Only users with a session open right now have their own counter; Slurm jobs are counted together, not per user.