A Centrality-Based Resource-Aware Microservice Orchestration in Cloud-to-Edge Continuum

Smart Cities and Industry 4.0 have opened new application scenarios posing new challenges to networking services. The continu- ously growing demand for high quality of service (QoS) and the intrin- sic geographical distribution of this novel class of applications require novel and tailored deployment solutions. Microservice applications place- ment in edge computing requires balancing resource constraints, net- work efficiency, and end-to-end latency constraints. This work presents Centrality-based Resource-Orchestrator System (CeROs), a novel Mixed Integer Linear Programming (MILP) model that introduces an objective function based on the structural node importance based on its posi- tion within the network’s topology, referred as node centrality, which promotes efficient resource usage across the network without directly optimizing latency or individual resource consumption. This approach ensures a generalizable and unbiased formulation that can be adjusted to various deployment scenarios. The experiments have been conducted over synthetic yet realistic cloud-to-edge network topologies and microservice applications, modeled to reflect real-world resource constraints, commu- nication delays, and deployment challenges. The evaluations demonstrate that CeROs outperforms trivial placement strategies, achieving compet- itive latency even in strict scenarios and more balanced resource usage, resulting in a lower node costs. CeROs does not force optimization toward a single metric, making it more robust across different emerging environ- ments as like smart cities with the Autonomous Vehicle Driving appli- cation.