How long does one device take to download an update?
Time is the size in bits divided by the speed you really get. The speed you really get is the link rate times an efficiency.
Example: a 1.5 MB image on a 300 kbps link at 60% efficiency is 12,000,000 bits ÷ 180,000 bit/s = 66.7 s. The same image at 26 kbps and 100% efficiency takes 461.5 s.
| Link | Rate (kbps) typical | Time per device |
|---|---|---|
| LTE Cat-1 | 10,000 | 2.0 s |
| LTE Cat-M1 | 300 | 1 min 7 s |
| NB-IoT | 26 | 12 min 49 s |
| 2G (GPRS/EDGE) | 40 | 8 min 20 s |
The Cat-1, Cat-M1 and NB-IoT rates are GSMA peak figures, and the 2G rate is typical. Your operator, signal and server decide the real number, so measure one device first and enter the result.
What is a staged rollout?
Update a few canary devices first, watch them, then widen the group: 5%, 25%, 50%, 100%. If devices fail more than your threshold, stop and roll back before the next wave.
| Stage | Devices | Total so far | Date (2026) | Soak days after |
|---|---|---|---|---|
| Canary group | 10 | 10 | 15 Nov | 2 |
| Wave 1 (5%) | 15 | 25 | 18 Nov | 2 |
| Wave 2 (25%) | 100 | 125 | 21 Nov | 2 |
| Wave 3 (50%) | 125 | 250 | 24 Nov | 2 |
| Wave 4 (100%) | 250 | 500 | 27 Nov | 0 |
The schedule assumes 100 devices update at once, 8 hours a day, each in about 1 minute, with 2 soak days after every step except the last, and no weekend rule. A slower link or a lower concurrency lengthens the waves.
How much SIM data does a rollout use?
Each device downloads the firmware and the config file, plus protocol overhead. Add the devices that retry, and keep a reserve for rollbacks.
| Item | How it is calculated | Result |
|---|---|---|
| One device | (1.5 MB + 20 KB) × 1.10 overhead | 1.67 MB |
| Fleet, no retries | 500 devices × one device | 836 MB |
| Fleet with 8% retries | Fleet × 1.08 | 903 MB |
| Rollback reserve | 10% of 500 devices × one device | 83.6 MB |
On a plan of 5 MB with 3 MB of normal use, a SIM has 2 MB spare. One download of 1.67 MB fits. A download plus a retry (3.34 MB) does not, so spread the waves over two months or raise the plan before you start.
What goes on a go / no-go checklist?
The planner calculates the first six checks from your inputs. A person must confirm the rest.
- Spare data on every SIM covers one download, and the pooled data covers every month.
- Wave percentages increase and the dates avoid your weekend days.
- A stop threshold is written down before wave 1.
- The previous firmware is stored and was installed on a bench device.
- A bench device survived a power cut during an update.
- Your platform shows the firmware version of every device after the update.
- One named person can stop the rollout and start a rollback.
What usually goes wrong?
- Slow devices stall the wave. A few devices in weak signal can take far longer than the average. Do not size the window on the average alone.
- Data runs out on small plans. A retry doubles the download for that SIM. Check one SIM, not only the pool.
- Devices are off when the wave runs. Vehicles that are parked for days miss a wave. Plan a catch-up pass for them.
- No way back. If the previous image is missing, a bad update cannot be undone. Store it and test it first.
Assumptions and limits
- Link rates, efficiency, overhead and the retry share are typical planning values. Replace them with your own measurements.
- The planner does not know the firmware size, delta-update support or resume behavior of any Tracom model. Enter the figures your device and platform really use.
- Every device is assumed to have its own SIM and to download the full package. A resumed or partial download uses less data.
- Waves run one after another. Your platform may start several at the same time or limit the hours per day.
- The checklist is a planning aid. It is not an approval, and the decision to go stays with your team.
- The calculations run in your browser. Your inputs are sent to our server, and kept for up to 45 days, only if you ask for the emailed report.
Sources
- Public OTA deploy guidance (canary and staged waves)
- GSMA: data throughput of LTE-M, NB-IoT, 3G and 4G
Last checked 3 November 2026. The 2G rate and the staged-wave practice have no official text here: measure your own network and verify before you rely on a figure.