Network Anomaly Event Prediction and Optimal Resource Control in Cloud Native Network Functions
Talkat Cloud Native Telco Day Europe 2022 · on Monday, 2022-05-16· at 13:25· Pavilion 3, Room E | Level 2 | Central Forum·
summary
eBPF per-container metrics are collected from 5G CNF containers and fed to an LSTM that predicts key KPIs such as UE registration failures, showing accurate prediction by ~150 seconds while basic cAdvisor metrics fail. An AI engine (SVR, Lasso, encoder‑decoder) plus a direct OSM interface with per-second cgroup CPU polling enables designated scaling and CPU-amount changes, supports NF migration, and runs on a 400 km JGN testbed; eBPF per-container metrics and LSTM prediction accurate at 150s.
AI-written summary · about
chapters (10)
- 0:00Intro
- 2:37Example of network anomaly events
- 5:06Overview of our approach
- 7:30Per-container detailed system metrics collected by eBPF
- 12:16Summary - Optimal Resource Control for CNFS
- 13:44Framework of Resource Operations for XNES
- 15:09Verification infrastructure on a public network testbed (GN)
- 16:16ETSI OSM based XNF operation system
- 19:43Benefits of Direct Interface and components
- 21:19General Conclusion
related talks
29mRuntime Security at Scale With EBPF: Hybrid Detection and Root Cause…
22mHarnessing Observability for 5G Performance: eBPF and OpenTelemetry I...
34mTroubleshooting Low Latency Application on CNF Deployment
42meBPF-Powered Observab...
35mSustainability Research the Cloud Native Way
19mNetwork X-Ray Vision: eBPF for Cloud-Native Observability
18mNoisy Neighbor Detection with eBPF
47mImprove Load Balancing With Machine Learning Techniques Based on the Sched_ext Framework