Optimizing CNF Placement for Latency and Resource Efficiency in Multi-Cluster Cloud–Edge Federations

Service placement in current and next-generation mobile networks in emergency management is increasingly challenging due to resource-constrained edge clusters, where low latency and efficient workload distribution are critical. Traditional service placement strategies that rely solely on resource availability are no longer adequate. Modern deployments must address complex constraints, including end-to-end latency, system load distribution, and dynamic resource availability. In this paper, we propose a comprehensive architecture for service orchestration across multi-cluster Cloud-Edge environments, where independent clusters coordinate placement decisions through a federation layer. Our system is built on the Karmada federation control plane, which unifies heterogeneous cloud and edge clusters under a common framework to support optimal service placement. The core contribution of this work is a service placement solution integrated into a multi-cluster service management platform, supported by a decision-making module for service prioritization and optimal allocation. The decision process has two main components: (1) a multi-objective optimization model that balances CPU utilization and latency through a weighting mechanism, and (2) an Integer Linear Programming (ILP)-based model that prioritizes services and selectively discards requests when edge resources are constrained or system requirements cannot be met. A metric aggregator module is deployed via agents in each cluster and continuously monitors resource status to support placement decisions. Simulation results confirm that our approach ensures efficient resource usage and low latency, while maintaining low execution time even for large problem sizes. Additionally, we experimentally evaluate the proposed solution using our federated multi-cloud 5G testbed.