FINE-STACK: Flexible architecture for low-power lossy IoT Networks with programmablE network STACK and semantic runtime updates

Low-Power and Lossy Wireless Networks (LLWNs) operate under stringent energy, memory, and bandwidth constraints and rely on inherently unreliable multi-hop links. Despite substantial progress in IoT networking, updating deployed LLWNs often still requires bulky, full-image firmware upgrades, even for minor network stack changes. This practice incurs excessive overhead, complicates reliable delivery over lossy links, and can temporarily disrupt connectivity due to mandatory reboots. We introduce FINE-STACK, a flexible architecture that decomposes the network stack into lightweight virtual functions coordinated by a central orchestrator. By enabling fine-grained composition and semantics-aware management of protocol components, FINE-STACK supports fast and selective over-the-air updates at runtime. It further improves the programmability of the network stack by enabling hybrid deployments that combine local in-network control with optional global optimization. We implemented and evaluated FINE-STACK on a real-world testbed, comparing programmability and performance across conventional deployment models and the hybrid model enabled by our design. The results demonstrate that hybrid deployments provide improved programmability and superior operational performance, while fine-grained modularity of the network stack substantially reduces update size and update time without violating strict memory, power and overhead constraints. Overall, FINE-STACK advances LLWNs toward practical network programmability, narrowing the gap between flexibility and efficiency, and making semantic runtime reconfiguration feasible on highly constrained IoT devices.