The problem
Operators inherit device estates that span vendors and firmware generations: OpenWrt on one SKU, RDK-B on another, a mesh vendor's cloud on a third. Each comes with its own management stack, its own tenancy model and its own way of bricking a device during an update. Consolidating onto one control plane only works if that plane keeps working at the scale of the whole fleet, on a bad day.
What we built
optimacs is the ACS behind the fleet figures on this site. It speaks TR-069 to the installed base and TR-369 / USP to anything newer, normalises both into the TR-181 data model, and keeps tenants separated by MSO so a single cluster can serve many operators without a per-operator deployment.
- Firmware campaigns are health-gated: 5% of the cohort first, then 25%, then the rest, each step gated on the KPIs the previous step reported.
- A campaign that regresses a KPI rolls back without a human in the loop; the rollback is journaled alongside the campaign.
- Device identity is per-device certificate, enrolled over EST, so the private key never leaves the hardware.
- On the device side, the ac-client agent gives OpenWrt access points a full USP 1.3 implementation with a UCI backend and hybrid post-quantum TLS.
How it holds
The control plane runs on the same Talos-on-bare-metal clusters we operate for our own services, fronted by AS207819, and reports 99.998% availability over the trailing 90 days. Scale testing has been run against a fleet of more than four million devices. The live dashboard on the home page is a fixture with sample telemetry; the fleet count and the availability figure are real.
Standalone edition
Not every operator wants to be onboarded onto a suite. The standalone edition is a single Rust binary with Postgres as its only dependency, packaged with published Helm charts, so an ISP can run its own ACS on its own hardware and still get the same protocol coverage.