AI-driven Orchestration at Scale: Estimating Service Metrics on National-Wide Testbeds

Network Slicing (NS) realization requires AI-native orchestration architectures to e fficiently and intelligently handle heterogeneous user requirements. To achieve this, network slicing is evolving towards a more user-centric digital transformation, focusing on architectures that incorporate native intelligence to enable self-managed connectivity in an integrated and isolated manner. However, these initiatives face the challenge of validating their results in production environments, particularly those utilizing ML-enabled orchestration, as they are often tested in local networks or laboratory simulations. This paper proposes a large-scale validation method using a network slicing prediction model to forecast latency using Deep Neural Networks (DNNs) and basic ML algorithms embedded within an NS architecture evaluated in real large-scale production testbeds. It measures and compares the performance of di fferent DNNs and ML algorithms, considering a distributed database application deployed as a network slice over two large-scale production testbeds. The investigation highlights how AI-based prediction models can enhance network slicing orchestration architectures and presents a seamless, production-ready validation method as an alternative to fully controlled simulations or laboratory setups. Keywords: Network Slicing, Deep Neural Networks, Machine Learning, Service-Level Agreement, Distributed Database1. Introduction Modern applications require challenging behaviors from physical networks to satisfy stringent requirements such as ultra-reliability, low latency, and high throughput [1]. In addition to these quantifiable network requirements, it is necessary to incorporate seamless, intelligent, and pervasive network capabilities to satisfy user demands [2, 3]. Although network management, control planes, and data planes have evolved to address this issue, challenges remain and require further large-scale evaluation. Many approaches, technologies, and methods have been developed to build user-oriented network architectures that provide connectivity in an isolated and personalized manner [4]. One key technological enabler of this vision is network slicing, which establishes network connectivity on top of physical infrastructure while ensuring isolation, end-to-end connectivity, and application-driven requirements, with dedicated control and data planes [5]. With this service-tailoring capability, Machine Learning (ML) e ffectively addresses various management and orchestration challenges, thereby enabling intelligent and real-time insights for service provider managers. AI techniques, such as reinforcement learning, supervised learning, and unsupervised learning, have been e ff ectively integrated with network orchestrators to mitigate cybersecurity threats, enable intelligent resource allocation, and ensure Service-Level Agreement (SLA) assurance for network slicing [7, 8, 9, 10].