Lightning Talk: eBPF-Powered Observability for Telco CNFs
Lightningat Cloud Native eBPF Day Europe 2022 · on Monday, 2022-05-16· at 15:25· Pavilion 4 | Room A | Level 2 | Central Forum·
summary
Carrier-grade observability for 5G CNFs uses eBPF (tools bcc and bpftrace) to collect fine‑grained TCP and process metrics every 10s, map them to CNF type via pid‑based and ip‑based mapping, and store JSON records for ML. Two failure scenarios (packet loss and CPU spike) run on a Kubernetes 5G core; LSTM models trained with eBPF data detect rising registration failures at 150 seconds with much higher accuracy than models without eBPF.
AI-written summary · about
chapters (7)
- 0:00Intro
- 0:23Proactive Operation for Carrier-grade Carrier-grade requirement
- 2:02Proposal Predictive failure detection using the ML model
- 3:12Implemented tools Our failure scenario case
- 4:07Mapping per CNF Raw data derived by bcc/bpftrace are not provided per CNF container Correlate data by key information pid-based mapping
- 7:20How much can eBPF contribute to prediction? Training
- 8:27Result Prediction performance at 150 sec from the event start
related talks
11mLightning Talk: 5G CNFs Observability Based on eBPF
36mFrom Monitoring to Observability: eBPF Chaos - Michael Friedrich
36mCloud Native Superpowers with eBPF
51mCloud Native Superpowers with eBPF
31meBPF for Observability: The Good, the Bad, and the Ugly
27mLow-Overhead Tracing Using eBPF for Observability into Kubernetes Apps and Services
42mCommunity Driven Breakout Session: eBPF and Auto-Instrumentation
40mPanel Discussion: eBPF: A New Era in Cloud Infrastructure Tools