Finding
Systems that keep improving after deployment follow six design rules; breaking any one caps what software can change
The same design rules show up in every successful software-defined system, from Tesla's vehicles to modern charger networks. Real-time control is kept separate from data traffic so a software update or video stream cannot delay a brake command. Hardware is abstracted so the software can move to a new chip. Computing is centralized in a few powerful computers, while inputs and outputs are handled by zonal controllers near the devices. Every update is signed, staged and reversible, and telemetry is designed in from the start.
These principles are easier to apply to a new design than to retrofit, which is why EV-native automakers and new depot operators move faster than incumbents with decades of supplier-specific controllers. See Why Full OTA Is Still Rare.
Principles
The six principles
| Principle | What it means | If ignored |
|---|---|---|
| Separate planes | Real-time control isolated from data traffic | Updates or video can disrupt safety functions |
| Abstract hardware | Software runs on a hardware abstraction layer | Every chip change means a rewrite |
| Centralize compute | Few high-performance computers | Dozens of ECUs to update and secure |
| Localize I/O | Zonal controllers near sensors and actuators | Heavy wiring, hard service |
| Update safely | Signing, staged rollout, A/B partitions, rollback | Bricked devices, failed recalls |
| Observe everything | Structured telemetry from day one | Blind operations; no learning loop |
Sources
Sources and scope
- Scope
- SDS Architecture Principles
- Sources
- ElectronsX framework
- Updated
- October 2026