SmartEdge P4-Driven Swarm Networking
SmartEdge’s introduces a P4-programmable network layer that automates device discovery, secure subnet (“swarm”) formation, and dynamic reconfiguration in industrial and edge environments. Recipe-based orchestration, combined with Access Point and Swarm Coordinator modules, enables seamless node onboarding, controlled or unexpected exits, and priority-aware data flows. Early tests confirm the solution’s efficiency, resilience, and real-world applicability.
Section 3 of the D4.2 deliverable describes the design and first implementation of SmartEdge’s automatic discovery and dynamic network swarm formation. It relies on P4-programmable data-plane capabilities to enable flexible, high-performance forwarding logic within smart factories and other use cases—such as connected vehicles or health monitors. By combining P4 switches with a set of coordinating software modules, SmartEdge orchestrates isolated, secure subnetworks—called swarms—where nodes discover each other, join or leave dynamically, and exchange data efficiently.
P4 Data Plane Programmability for Swarm Formation
The core idea is to use the P4 language to define forwarding behaviors directly in network devices. SmartEdge applies P4 across both the network and data-plane layers of components—mobile robots, storage racks, IoT gateways—that act as Swarm Smart Nodes or as Access Points and Swarm Coordinators. Each swarm operates in its own P4-configured subnet, ensuring traffic isolation and meeting strict latency requirements. Optional features such as in-band telemetry and 1+1 protection can also be embedded in the P4 switch for monitoring and reliability.
Key Components
- Swarm Smart Nodes run the Smart Node Manager software to expose their local network state and manage P4 switch rules.
- Access Points host a P4 switch plus the Access Point Manager, which handles device authentication (e.g., Wi-Fi), IP assignment, and VXLAN tunnel setup.
- Swarm Coordinators maintain each swarm’s address space, orchestrate node onboarding/offboarding, and store node state in a distributed database (the ART).
Swarm Discovery and Formation
When a new smart node enters an AP’s coverage area, it first performs standard Wi-Fi authentication. The Access Point Manager then assigns it an IP address, sets up a VXLAN tunnel for control traffic, and records its metadata (IP, MAC, UUID) in the ART. Initially, P4 rules on the AP allow this node to speak only with swarm coordinators. The node learns which swarms are available, sends a join request to a coordinator, and—once accepted—the coordinator updates the ART and deploys new P4 rules on all relevant APs to permit full intra-swarm communication.
Recipe-Based Orchestration
SmartEdge also supports automated swarm formation based on recipes. A Task Orchestrator submits a recipe (defining application requirements and required node capabilities) to a Swarm Coordinator. The coordinator queries the Thing Description Directory (TDD) for matching, available nodes, onboards them, and returns the selected list. The orchestrator validates these nodes against Industrial Knowledge Graphs and, upon success, assigns application tasks. This ensures that each swarm precisely meets functional and performance needs.
Controlled and Unexpected Exits
Nodes can leave a swarm gracefully by sending a leave request to the coordinator, or the coordinator can remove them when they are no longer needed. In both cases, the coordinator deletes the node’s record from the ART and updates P4 rules on APs to block further traffic. For unexpected disconnections—detected via Wi-Fi disconnect events or missing heartbeats—the AP waits a configurable timeout, then purges routing entries and marks the node as absent in the ART. These mechanisms combine to deliver automatic recovery and maintain swarm resilience.
Catering to Diverse Data Sources
In addition to SmartEdge-compliant nodes, swarms may include legacy or non-smart devices (cameras, LiDARs, humidity sensors) that provide vital data. The P4 layer dynamically recognizes traffic from these sources, updates the swarm state, and assigns priorities. For example, camera feeds for location tracking receive higher priority than environmental readings, ensuring timely obstacle detection and safe operation of mobile robots.
SmartEdge Network Layer Architecture
Each swarm node comprises two layers: a P4 switch and a Network Control Plane (CP). The P4 switch enforces forwarding rules, handles VXLAN encapsulation, and injects or extracts in-band telemetry. The CP layer manages swarm formation and updates P4 match-action rules. This separation of data forwarding (P4) and control logic enables rapid reconfiguration and fine-grained monitoring of node metrics such as latency and packet flows.
Swarm Networking Artifacts
SmartEdge defines five key artifacts for swarm networking:
- A4.1 Access Point Manager—runs on APs to create VXLAN interfaces, assign IP addresses, and install P4 flow rules.
- A4.2 Swarm Coordinator—manages P4 switches, enforces access control, and maintains the ART.
- A4.3 Swarm Node Manager—runs on smart nodes to establish tunnels, monitor connectivity, and handle join/leave actions.
- A4.4 P4 Switch—enforces intra-swarm forwarding, isolation, and in-band telemetry on APs and nodes.
- A4.5 ART (Distributed Database)—stores each node’s UUID, swarm ID, IP/MAC, status, timestamp, and RSA signature for secure, auditable management
Preliminary Results
A series of 15 test runs measured performance at the Access Point, Swarm Coordinator, and Smart Node, focusing on system efficiency, bottleneck identification, and operational behavior under varying conditions. While detailed metrics are documented in the full deliverable, early findings confirm the feasibility, responsiveness, and resilience of the P4-based swarm networking design, paving the way for further optimization and real-world deployments.