Lab load
How busy each board type in the lab is, and how long jobs wait for a device. Every job on a board counts: CI, health checks, manual and other submitters, because they all occupy the devices.
Data as of 29 Oct 2026, 09:00 UTCLast synced:LAVA 29 Oct 2026, 09:00 UTCJenkins 29 Oct 2026, 09:01 UTCS3 export 7 Oct 2026, 09:53 UTC
Window
Show
Where demand outstrips the lab
Board types over the last 12 months, longest p90 queue wait first.
Status follows the p90 queue wait over the window: Busy at an hour or more, Queueing at 15 minutes or more. A board whose devices are busy 70% of the time or more is Near capacity even while waits are short. Otherwise it is Keeping up. Long waits with low utilisation usually mean devices were offline rather than too few.
Jobs submitted per board type and week
Scroll the heatmap sideways to see every week.
| Board | |||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Other | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Juno-R2 | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Nucleo-U3C5ZI-Q | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Arm | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Juno | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Musca-B | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| N1SDP | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Chromebook | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Asurada (MT8192) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Cherry (MT8195) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Corsola (MT8186) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Geralt (MT8188) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Rauru (MT8196) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Trogdor (SC7180) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| NXP | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| FRDMMCXN947 | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| LPCXpresso55S69 | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| Raspberry Pi | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| RPi (BCM2837) | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| STMicroelectronics | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| B-U585I-IOT02A | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| STM32H573I-DK | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| STM32MP15X-Eval | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| STM32U3C5ZI-Q | |||||||||||||||||||||||||||||||||||||||||||||||||||||
| STM32WBA65I-DK1 | |||||||||||||||||||||||||||||||||||||||||||||||||||||
- 1 to 24
- 25 to 99
- 100 to 499
- 500 to 1,999
- 2,000 or more
- 0 submitted; a device was still busy with earlier jobs
- No jobs
- No data held
How these are measured
- Utilisation
- Time devices spent running jobs, cut at the edges of each day or week, divided by the devices that ran any job on that board in that day or week times its length. A device that ran nothing in a bucket is not counted, so spare idle units do not lower the figure. The window figure divides all busy time by the summed capacity of the cells. Days without data add neither busy time nor capacity, and the last day runs only to the data's as-of time.
- Health-check share
- The part of the busy time spent on LAVA health checks rather than test jobs.
- Queue wait
- Time from submission to start for test jobs submitted in the bucket that started on a device. Health checks are left out because LAVA schedules them ahead of the queue, and jobs cancelled before they started are left out. p50 is the median; p90 is the wait that 9 in 10 jobs beat (nearest rank). Waits also grow while devices are offline.
- Jobs
- Every job submitted for the board type in the bucket, health checks included.
- Scope
- Lab boards only: virtual targets such as FVP and QEMU are left out, as are retired boards with no jobs in the window. Only finished jobs are held, so jobs still queued or running at the last sync are not counted.