AI Infrastructure (elective) - Map of Content
Success: Elective complete
All four AI Infrastructure sections (1-4) are written - your CCDE Practical Scenario 4. Notes are numbered
section.NNin study order. The flagship is Fabric transport - Ethernet vs InfiniBand vs UEC.
1.0 AI/ML fundamentals & governance
- AI-ML workloads and their infrastructure impact
- Service placement - on-prem, cloud, hybrid, distributed
- Data sovereignty and data gravity
- Regulations, governance, and AI security policy
- Cost, ROI, and scalability of AI infrastructure
- Workload placement, mobility, and vendor lock-in
2.0 Network
- RDMA, RoCE and RoCEv2
- Fabric transport - Ethernet vs InfiniBand vs UEC
- Lossless fabric - PFC, ECN, DCQCN
- Rail-optimized CLOS topology
- Optics and cabling selection
- Latency and the straggler problem
- Bandwidth and capacity planning
- PTP timing for AI fabrics
- Connectivity models and SDN in AI fabrics
- Validated reference design - Cisco AI-ML lossless fabric (CVD)
- Parallelism strategies and collective traffic patterns
- Load balancing in AI fabrics - flow, flowlet, packet spray
- Fabric resiliency and failure handling
3.0 Security
- AI security design and hosting models - air-gapped, private, public
- Securing the high-performance fabric
- Tenant isolation and segmentation in shared AI clusters
- Model and data protection
- Protecting AI services - WAF, API security, and inference endpoints
- AI-assisted threat detection (AI for security operations)
4.0 Compute, storage & environment
- AI-enabling hardware - GPU, DPU, SmartNIC
- Scale-up vs scale-out
- Storage and checkpointing for AI
- Sustainability - Green AI, power and cooling
- AI-assisted operations - AIOps
- Storage networking - FC, FCoE, iSCSI, and NVMe-oF
See also: CCDE Core - MOC.