RMM capacity calculator
Tell us how many devices you manage and how often they report. We show the load, the storage and a server layout to start from.
You will need a separate database server at this size. Keep the database off the app server, with the disk and memory shown below.
At this size the app and the database can share one VM. A separate database server becomes necessary from about 2,000 devices or 300 rows per second.
The tier is the larger of your device count and your rows per second. These layouts are our own estimates and have not been load-tested.
Load
- Check-ins per second
- Rows per second
- Statements per second
- PHP workers busy
- Rows per check-in
- Statements per check-in
Storage
- Raw samples per month
Extended metrics store about as much as the basic set for this fleet. Turn them off above to see the difference; the default is much bigger than the table most people remember.
What to set
- PHP-FPM
pm.max_children - Database disk
- Ingest
Twice the steady storage plus 20 GB, for indexes and the other tables.
How we calculate this
Check-ins per second is devices ÷ check-in interval. Each check-in sends max(1, round(interval ÷ sampling)) batches. A batch holds 7 samples, or 7 + 13 + 2 × disks + cores with extended metrics. A check-in writes 7 rows plus its samples and runs 12 statements, plus 7 and one per batch when metrics are on.
Four constants are design assumptions, not measurements: 7 rows per batch, 12 base statements, 120 bytes per row and 0.15 seconds per request. With 10,000 devices, 300 second intervals and extended metrics off, the model gives 33.3 check-ins/s, 467 rows/s, 667 statements/s, 5 busy workers (15 at peak), 73 GB a month and 17 GB at 7 days, which matches the table in the capacity doc.
The tiers and the derived values are our estimates, not taken from the doc. Past 5,000 devices they are design estimates only. MeshCentral keeps one connection per device and is sized separately.