Cost-effective reliability allocation for Software-Defined Networks based on an application-oriented approach

Ensuring application-level reliability in Software-Defined Networking (SDN) systems is essential for critical infrastructure services. However, current reliability allocation methods mostly focus on physical-layer metrics (e.g., MTBF) and path-level redundancy, overlooking the quantifiable influence of SDN’s logical-layer mechanisms, such as the controller fault detection rate and restoration latency on end-to-end application reliability. This paper addresses this gap by proposing an application-centric reliability allocation framework tailored to SDN design-time planning. We first develop a multi-layer reliability modeling architecture that captures the coupling between application application-level objectives and both physical and logical component parameters. Then, we formulate a cost-minimization model with strict reliability constraints and solve it using an evolutionary optimization algorithm customized for SDN control-data plane dynamics. Experimental results on two real-world SDN cases show our approach achieves equivalent application reliability at 33%–75% lower cost compared to traditional redundancy-based schemes. Further analysis reveals that tuning logical parameters without upgrading hardware can compensate for up to 60% of physical reliability improvements, enabling cost-efficient design strategies for diverse application demands. Our method enhances the flexibility of SDN reliability design and offers generalizable insights into cross-layer reliability optimization for software-defined systems.