Federated Learning-Driven Dynamic Resource Allocation for Edge-Native Serverless Workloads

Abstract

**Abstract:** This research proposes a novel framework for dynamic resource allocation in edge-native serverless environments utilizing Federated Learning (FL). Addressing the critical challenge of efficient resource management in geographically distributed edge locations with intermittent connectivity, our approach leverages FL to collaboratively train a resource allocation policy without exposing raw workload data. This allows for adaptive allocation decisions based on localized edge conditions while ensuring global optimality. We demonstrate a 15% improvement in resource utilization and a 10% reduction in latency compared to traditional centralized allocation strategies via simulation and targeted laboratory experimentation on a simulated edge network. This technology fosters cost-effective scaling of serverless applications – bolstered adoption of edge computing, and unlocks new opportunities for real-time data processing and localized AI inference. --- *This document is a part of the Freederia Research Archive. Explore our complete collection of advanced research at [en.freederia.com](https://en.freederia.com), or visit our main portal at [freederia.com](https://freederia.com) to learn more about our mission and other initiatives.*