Cluster Status
Live statistics · updated 47 seconds ago · refreshes automatically
Running jobs
90
right now
Queued jobs
1,479
waiting to start
GPUs in use
77%
79 of 102 available
Jobs completed
1,476
in the last 24 hours
Cluster health
- Compute nodes available
- 22 / 27
- GPUs available
- 102 / 111
- GPUs free right now
- 23
- Scheduler
- responding · 2.0 ms cycle
Nodes by state
In use 21Idle 1Drained 2Down 3
Last 24 hours
- Jobs started
- 1,495
- Jobs completed
- 1,476
- GPU hours delivered
- 1,800
- Active users
- 40
- Active projects
- 29
Partitions
all_serial
Operational- Running / queued
- 4 / 0
- GPUs in use
- 4 / 24 (17%)
- Nodes available
- 2 / 2
Wait time · last 7 days
| Median | 90th pct | |
|---|---|---|
| First job (39) | 0 min | 0 min |
| With other jobs active (80) | 0 min | < 1 min |
all_usr_prod
Operational- Running / queued
- 35 / 577
- GPUs in use
- 75 / 78 (96%)
- Nodes available
- 20 / 25
Wait time · last 7 days
| Median | 90th pct | |
|---|---|---|
| First job (239) | < 1 min | 3h 30m |
| With other jobs active (2,507) | 14 min | 13h 31m |
boost_usr_prod
Operational- Running / queued
- 51 / 902
- GPUs in use
- 54 / 54 (100%)
- Nodes available
- 14 / 14
Wait time · last 7 days
| Median | 90th pct | |
|---|---|---|
| First job (113) | 32 min | 2h 54m |
| With other jobs active (2,099) | 2h 22m | 21h 39m |
Last 7 days
GPUs in use
Queued jobs by partition
- Wait time is measured from when a job becomes eligible to run (after any dependency or requested start time) to when it starts, for jobs that started in the last 7 days. The number in brackets is how many jobs each figure is based on.
- First job: the user had no other job running or queued when this one became eligible to run. With other jobs active: the user already had jobs running or waiting. Scheduling priority decreases with recent usage, so jobs from users who are already running tend to wait longer.
- Partitions share nodes:
boost_usr_prodnodes are also part ofall_usr_prod, and the same free GPU can appear in both. Cluster totals count each node once. - A GPU is available if its node is up and accepting jobs; nodes that are drained, down or in maintenance are excluded.