Advanced Networking
Network operations are now data- and agent-mediated: telemetry is triaged by machine-learning detectors, configuration and remediation are drafted by AI tooling that must be verified before it reaches production, and AI workloads themselves have become a dominant traffic class shaping fabric design. The course keeps its networking identity and adds the measurement, detector-evaluation, and control-plane accountability skills that shift demands.
Current description → proposed description
This is an advanced course which focuses on modern trends in computer networking technology. While this course will be related to the other networking course in this curriculum, it takes a different approach. Focus is placed on advanced topics related to emerging computer networking concepts.
This advanced course examines modern trends in computer networking technology and the operational practices that now surround them. Related to the other networking course in the curriculum, CSC 5040 Applied Computer Networking, it takes a different approach, concentrating on emerging architectures and on the measurement, automation, and defense of networks at scale. Topics include software-defined and intent-based networking, overlay and segment routing, modern transport protocols such as QUIC and multipath TCP, multi-tenant data-center fabric design, network function virtualization, edge and mobile network integration, and the traffic characteristics of distributed model training and large-scale inference serving. Because network operations are increasingly mediated by machine learning and by tool-calling agents, the course also treats streaming telemetry pipelines, machine-learning traffic classification and anomaly detection and their evaluation at realistic base rates, the verification of machine-generated configuration before deployment, and the exposure that agentic tooling introduces into the control plane through prompt injection and tool supply-chain risk. Students weigh the privacy and accountability consequences of automated traffic inspection and communicate design rationale and incident findings to technical and non-technical audiences.
What changes
- Names the advanced topics the current description leaves entirely unspecified
- Differentiates the course from CSC 5040 by content rather than by assertion
- Adds streaming telemetry pipelines and operational evaluation of ML traffic classification and anomaly detection
- Adds verification, rollback, and prompt-injection and supply-chain threat modeling for agentic control-plane automation
- Treats AI workload traffic, training collectives and inference serving, as a first-class fabric design driver
5 proposed outcomes, mapped to 5 program outcomes
Each outcome below is written to be observable and assessable, and each is mapped to the program learning outcomes for which it produces evidence.
Students will be able to analyze the behavior of modern transport, routing, and overlay mechanisms, including QUIC, multipath TCP, congestion-control variants, and segment routing, under loss, latency, and contention, by collecting and reconciling packet captures, flow records, and streaming telemetry drawn from device APIs and time-series stores.
The CLO's evidence is the collection, cleaning, and reconciliation of measurement data from several distinct sources, packet captures, flow records, and telemetry retrieved through device APIs and stored in time-series databases, in order to answer a real operational question, which is the multi-source data acquisition and analysis PLO 2.1 names.
Students will be able to design a multi-tenant network architecture and its programmable control plane for a bandwidth- and latency-sensitive workload, such as a data-center fabric carrying distributed model training collectives alongside high-volume inference serving or an edge deployment serving latency-bound inference across mobile access networks, sizing topology, congestion control, and segmentation from traffic characteristics the student derives by profiling collected flow records and telemetry for that workload, and justifying virtualized function placement, policy distribution, and failure-domain boundaries against explicit latency, loss, and cost budgets.
The PLO 2.1 evidence here is acquisition the student performs rather than receives, profiling collected flow records and telemetry to derive the workload's traffic characteristics, which are then what size the topology, congestion control, segmentation, and virtualized function placement of a real deployment; this differs from CLO 1, where multi-source measurement is reconciled to explain protocol behavior rather than to produce a workload profile that drives a design.
Students will be able to evaluate a machine-learning traffic-classification or anomaly-detection system in operation, judging precision and recall at realistic base rates, alert volume and false-positive cost, the limits imposed by encrypted traffic, and symptoms of drift, specifying a threshold-refinement and retraining procedure while weighing the privacy intrusion that the required inspection imposes on the users whose traffic is examined.
Judging the detector against precision, recall, false-positive cost, and drift and then specifying threshold refinement and retraining is exactly the metric-based assessment of a model's performance and limitations and the iterative refinement from diagnostic results in PLO 6.2; weighing the privacy intrusion that traffic inspection imposes on the monitored users is the privacy and potential-harm clause of the outcome the workbook states twice, word for word, under Christian Faith as PLO 1.1 and under Integrated Disciplinary Knowledge as PLO 3.2, so the single piece of evidence supports both ids.
Students will be able to critique an AI-assisted or intent-based network automation pipeline, specifying the verification a machine-generated configuration must pass before deployment, the staged rollout and rollback controls that bound its blast radius, the prompt-injection and tool supply-chain exposure it introduces into the control plane, and the point at which a named human becomes accountable for a change no person authored.
Fixing where accountability rests for a production change no person authored, and bounding the harm such a change can do through pre-deployment verification, staged rollout, and rollback, is the accountability and potential-harm clause of the outcome the workbook states twice, word for word, under Christian Faith as PLO 1.1 and under Integrated Disciplinary Knowledge as PLO 3.2.
Students will be able to communicate network design rationale, measurement evidence, and post-incident findings to both engineering and non-specialist stakeholders, stating plainly what automated detection or agentic tooling contributed, how its output was verified, and what remains uncertain.
Reporting what an automated detector or agent contributed, how it was verified, and what is still uncertain, to engineers and non-specialists alike, is the clear and responsible communication of AI results and their consequences to technical and non-technical audiences that PLO 5.1 names.
Program outcomes this course reaches
Filled cells are program learning outcomes with at least one supporting course learning outcome in this course. Sparse coverage is expected — no single course carries all twelve.