Training / Course 03
Stop watching the network and start changing it. Cut an end-to-end network slice, differentiate QoS with Linux traffic control and prove the shaping in the data plane, harden the core against the TS 33.501 threat model (SUCI/ECIES subscriber-identity protection included), reconstruct signalling from the NF discovery API, and stand up a KPI pipeline that feeds an AI root-cause on-ramp.
Aims — what you can do after
- Harden the core to TS 33.501: order the security algorithms and enable identity concealment with your own key pair.
- Engineer network slices end to end, and diagnose rejected or mis-routed slices.
- Differentiate QoS with subscription rate limits, policy rules and per-slice traffic shaping.
- Perform signalling forensics: query the NF discovery API and decompose a registration at the message level.
- Investigate mobility and session resilience — what core state survives a handover or teardown.
- Engineer a KPI pipeline: live exporters, a fast scrape loop, standardized metrics.
- Understand how a digital-twin root-cause agent automates the discipline you practised by hand.
The slicing chain — every link must agree
A slice is only real when the identifier at the AMF, the data network at the SMF and the subnet at the UPF all line up. Diagnose a rejected slice by finding the broken link.
Three layers where QoS is actually enforced
From telemetry to root cause
The engineering feeds a loop: the KPIs and labelled scenarios you produce become the training data for a digital-twin root-cause agent — and where an engineer's judgement still beats automation.
Module breakdown
What engineering the core looks like
- Deploy a network slice end to end and diagnose why one is rejected.
- Distinguish signalled QoS from enforced QoS and reconcile them.
- Read the service-based architecture itself for the faults that have no obvious layer.