Abstract
Container orchestration platforms often utilize static resource requests and topological labels to schedule workloads. The disclosed technology introduces a closed-loop control mechanism that dynamically biases workload placement based on real-time physical network health. The system acts as a normalization layer between a high-frequency data plane and a low-frequency control plane. By employing passive kernel-level telemetry, the system captures network state metrics from existing application traffic. Orchestrator control plane metadata is synchronized into the data plane, allowing network flows to be aggregated by physical failure domains. An adaptive normalization logic component differentiates between application-layer processing delays and network-layer transport delays to prevent false positive congestion detection. A scheduling component treats a normalized health score as a soft constraint, facilitating risk-adjusted placement without blocking scheduling operations. Keywords: container orchestration, network-aware scheduling, passive telemetry, closed-loop control, failure domain aggregation, differential latency analysis.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Ojha, Ankita, "System and Method for Closed-Loop Network-Aware Container Scheduling via Passive Kernel-Level Telemetry and Topology-Correlated Congestion Detection", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9611