Deliverable D3.2: Cloud continuum, service mesh and resource orchestration

This document represents the final software outcomes and architectural implementations of Task 3.1, reporting on the design, execution, and deployment of the Edge-Cloud continuum layer as well as the DataOps module for the SAFE-6G framework. The involved components compose the overall framework that manages distributed computing resources and orchestrates the lifecycle of microservices across heterogeneous domains, also monitoring their status and performance. At the core of the continuum implementation is a Meta-Operating System based on ECLIPSE aeriOS. This system offers a programmable environment designed to integrate seamlessly with standard Cloud Native technologies. It successfully enables the federation of delocalized computing domains and provides continuous observability, identity management, and distributed service orchestration using Kubernetes as its foundation. The Meta-OS relies on critical components including a FIWARE Orion-LD Context Broker for sharing context data without replication, and an enhanced dual-level orchestration engine to automate the deployment, upgrade, and removal of services packaged as Helm charts or containers. The flexibility of its data model has allowed the extension of its default definition with SAFE-6G specific entities and attributes. Also, the exposure capabilities of the continuum have been secured by an API Gateway built on KrakenD, fully integrated with OpenCAPIF Release 3 to facilitate automated API discoverability, access management, and interaction with the broader SAFE-6G Trust Functions. To complement the orchestration logic, a Multi-Domain Service Mesh has been successfully deployed utilizing Cilium. This module manages intra- and inter-domain connectivity, acting as a Cloud Native CNI plugin that enforces granular network policies. Thanks to the introduction of bidirectional VPNs and Cluster mesh capabilities, this layer guarantees secure service-to-service communications, comprehensive traffic observability, and multi-domain network security across the delocalized microservices. Finally, the document details the DataOps Implementation, the module of the SAFE-6G framework responsible for the telemetry, monitoring, and long-term storage of infrastructure metrics. Specifically, the DataOps deploys a robust Prometheus stack, integrated with Loki for domain-level log aggregation and Grafana for visual dashboards. To implement a complete observability, widely-used metrics exporters have been complemented by a set of custom ones, such as the Data Fabric exporter, 5G Core exporter, Metaverse manager exporter, and gNB performance exporter. Together, these modules generate the historic and real-time data strictly required to train models in MLOps pipelines and empower the project’s Trust Functions. This report summarizes the requirements considered for their realization and the features implemented, pointing to relevant documentation and with extended details on those components explicitly developed or adapted for the needs of SAFE-6G; this is, including additional information such as developed/available APIs, container images and Helm packages’ routes, required resources, etc.