As Internet of Things (IoT) applications become increasingly distributed and latency-sensitive, cloud-centric serverless architectures show limitations in constrained and heterogeneous edge environments, where device-level sensing and actuation increase infrastructure management complexity. This paper presents a modular and lightweight Function-as-a-Service (FaaS) architecture that extends serverless execution from the cloud to the edge, building on the Stack4Things (S4T) framework and Kubernetes-native orchestration principles. The proposed design supports a deviceless operational model in which users deploy and invoke functions according to capabilities, location, and workload conditions, without directly managing the underlying IoT devices. The architecture is organized into three tiers: a user-interface layer, centralized management nodes, and edge workers where functions execute close to data sources and actuators. It integrates Nuclio as a high-performance serverless runtime, a cloud-side load balancer for adaptive function placement, and WebSocket Tunnel WSTUN-based connectivity for devices behind NATs or firewalls. The resulting platform supports both Kubernetes and Docker-based deployments, improving portability across cloud, edge, and resource-constrained IoT nodes while preserving lightweight execution. Experimental evaluation on constrained single-board devices and cloud nodes shows limited resource overhead, balanced function distribution, and end-to-end latency below 300 ms. Under intensive workloads, the proposed solution also outperforms TinyFaaS, confirming its suitability for dynamic, low-latency IoT-edge serverless scenarios.
Toward Deviceless Paradigm: An IoT-centered and Orchestrated FaaS Architecture
Mirto F. O.;D'Agati L.;Tricomi Giuseppe;Longo Francesco;Merlino Giovanni
2026-01-01
Abstract
As Internet of Things (IoT) applications become increasingly distributed and latency-sensitive, cloud-centric serverless architectures show limitations in constrained and heterogeneous edge environments, where device-level sensing and actuation increase infrastructure management complexity. This paper presents a modular and lightweight Function-as-a-Service (FaaS) architecture that extends serverless execution from the cloud to the edge, building on the Stack4Things (S4T) framework and Kubernetes-native orchestration principles. The proposed design supports a deviceless operational model in which users deploy and invoke functions according to capabilities, location, and workload conditions, without directly managing the underlying IoT devices. The architecture is organized into three tiers: a user-interface layer, centralized management nodes, and edge workers where functions execute close to data sources and actuators. It integrates Nuclio as a high-performance serverless runtime, a cloud-side load balancer for adaptive function placement, and WebSocket Tunnel WSTUN-based connectivity for devices behind NATs or firewalls. The resulting platform supports both Kubernetes and Docker-based deployments, improving portability across cloud, edge, and resource-constrained IoT nodes while preserving lightweight execution. Experimental evaluation on constrained single-board devices and cloud nodes shows limited resource overhead, balanced function distribution, and end-to-end latency below 300 ms. Under intensive workloads, the proposed solution also outperforms TinyFaaS, confirming its suitability for dynamic, low-latency IoT-edge serverless scenarios.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


