Onsite ANPR Reliability

Context

• The ANPR workloads ran onsite in constrained infrastructure with intermittent network links.
• Operations teams needed predictable behavior during updates and during degraded connectivity.

Challenge

• Runtime drift across edge nodes caused inconsistent recognition behavior.
• Manual deployment steps increased release risk.
• Limited health visibility delayed incident detection.

Key Decisions

• Standardize edge runtime packaging. Why: Required less node-level drift and more consistent releases.
• Introduce health and deployment verification gates. Why: Required early checks to catch faults before full rollout.
• Add lightweight telemetry focused on failure modes. Why: Required clearer operational visibility without overloading constrained systems.

Implementation Changes

• Standardized runtime and deployment structure for edge nodes.
• Added health checks and release verification before rollout completion.
• Implemented telemetry for failed reads and processing latency trends.

Results

• Improved detection consistency in onsite processing windows.
• Faster identification of degraded nodes.
• Safer rollout process with a clearer rollback path.

Tech Stack