Skip to main content

Network Topology

A Smart Corridor node sits between three network concerns: the camera network feeding it video, the client network consuming its /graphql endpoint, and — in multi-node deployments — the inter-node link to the VPP master. Internal traffic between the Hub, VPP, CIGS, CBS, and the message broker stays inside the node's Docker network and never needs to be exposed.

Camera network

Cameras belong on a dedicated VLAN or physically separate segment reachable only by the node host. For server-side RTSP, budget roughly 4 Mbps of stable bandwidth per camera and keep latency low enough to hold 15+ fps end to end; PoE switches simplify cabling and power. On-edge devices send only metadata and selected frames, so their bandwidth needs are a fraction of RTSP — a practical relief on sites where pulling new cabling is expensive.

Client and integration network

External consumers — the officer and traveler UIs, integrator systems, gate controllers — need exactly one thing: HTTP(S) access to the Hub's /graphql endpoint (WebSocket or SSE for subscriptions). No client ever connects to VPP, the broker, or the database directly. Terminate TLS in front of the endpoint and restrict access to the operational network; see Security Hardening.

Inter-node and outbound

Multi-node deployments need connectivity between nodes and the VPP master (see Deployment Layouts); matching runs locally, so this link tolerates latency and interruptions without stopping lanes. Outbound access is needed at install and upgrade time to the Innovatrics and GitLab registries (see First Deployment); air-gapped operation between upgrades is otherwise viable.