Edge Computing for Real-Time IoT Monitoring
Why device-local processing was necessary for monitoring smart-city infrastructure at scale under intermittent connectivity.
Summary
This ADR covers the decision to process telemetry at the edge rather than relying solely on a centralized cloud pipeline for a smart-city IoT deployment.
Problem
Field devices — traffic signals, environmental sensors, waste-management systems — were distributed across many locations with connectivity that could not be assumed to be constant. A cloud-only design would lose data during any network interruption.
Decision
Process and buffer telemetry locally at edge gateways, syncing to the cloud platform when connectivity allows, rather than requiring devices to stream continuously to a central endpoint.
flowchart LR
Sensor[Field Sensor] --> Gateway[Edge Gateway]
Gateway -->|buffers during outages| Gateway
Gateway -->|sync when online| Cloud[Cloud Platform]Trade-offs
| Option | Pros | Cons |
|---|---|---|
| Cloud-only ingestion | Simpler architecture | Data loss during connectivity gaps |
| Edge processing (chosen) | Resilient to intermittent connectivity | More components to secure and operate |
Lessons Learned
Designing for network unreliability from the outset was far cheaper than retrofitting buffering and replay logic after a data-loss incident. The edge layer became the natural place to enforce device authentication and encryption as well, since it was already a trust boundary.