LIS: Latency-and-Interference-Aware Scheduling for Cloud-Native Network Functions
Shunbin Dong, Jin Zhao · 2025
The growing demand for reliable networking and the evolution of cloud technologies have driven the adoption of cloud-native network functions (CNFs). In resource-constrained environments such as Space-Air-Ground Integrated Networks (SAGIN), the co-location of CNFs on limited nodes can lead to significant performance degradation due to interference. However, existing cloud-native schedulers typically overlook this issue, lacking the ability to anticipate CNF-induced latency under constrained resources. Therefore, we propose a Latency-and-Interference-aware Scheduling (LIS) framework, a novel CNF scheduling framework for heterogeneous cloud-native environments. LIS models inter-CNF interference and incorporates this into a Kubernetes-based scheduling system, combining algorithmic design and system implementation. We conduct evaluations in a real-world deployment using actual network function workloads. Experimental results show that our framework improves overall network performance by up to $20 \%$ while maintaining low scheduling latency.