Distributed serverless function scheduling in ad-hoc drone networks

Abstract

The increasing use of Unmanned Aerial Vehicles (UAVs) in critical applications, such as disaster response, compels efficient communication and computation frameworks for highly dynamic ad-hoc networks. We present an interpretation of Function-as-a-Service serverless computing within the distributed settings of drone swarms to address their peculiar challenges in functionality deployment, resource allocation, and mission adaptability. We propose a novel two-layer network overlay architecture, combining a gossip-based topology management layer with a function scheduling layer to support distributed function scheduling. Our system introduces a declarative language of Ad-Hoc Allocation Priority Policies (AHAPP), tailored for ad-hoc drone networks, enabling flexible function deployment based on resource constraints and operational needs. The resulting combination addresses the volatility of UAVs networks by supporting execution semantics for stable- and dynamic-topology scenarios, function offloading, and resilience to network disruptions. We present experiments confirming that the features provided by our proposal support the efficient execution of serverless functions in ad-hoc drone networks, effectively handling their dynamic and heterogeneous nature, while achieving strong performance in terms of reliability, scheduling time, and communication overhead.