C01
Datacenter Fundamentals & Topologies
DC design goals, traffic patterns and the shift from 3-tier to spine-leaf.
Track C10 modules · 70 production-grade assets
Datacenter (AI Infrastructure Core)
The flagship track — spine-leaf fabrics, EVPN-VXLAN and lossless GPU-cluster networking.
10 modules, sequenced foundation → design 7 assets per module — theory always ships with a lab Career ladder with salary bands at every rung
What this track covers
One focus. One measurable outcome.
Focus
The whole thesis in one place: spine-leaf/Clos fabrics, VXLAN/EVPN, lossless Ethernet (RoCEv2/PFC/ECN/DCQCN) and AI/ML GPU-cluster networking.
Graduate outcome
Graduate can build and operate a lossless GPU-cluster fabric — the exact scarce skill behind an AI-operations Supply Sufficiency Index of 47 (one qualified candidate for every two roles).
Syllabus
10 modules, in the order you'll master them.
Each module is a complete unit of competence — sequenced so every later module stands on the one before it, and every one ships the same seven production-grade assets.
C02
Spine-Leaf / Clos Fabric Architecture
Design and build modern Clos fabrics: non-blocking theory, oversubscription math, ECMP and eBGP underlay design.
Module outcomes
- Explain why 3-tier STP designs fail under east-west traffic and state the Clos non-blocking theorem
- Calculate oversubscription ratios and size a 1:1 non-blocking fabric from server count and port specs
- Explain ECMP hashing, diagnose polarization and configure eBGP unnumbered underlays across vendor CLIs
- Connect Clos/oversubscription theory to AI/GPU cluster fabric requirements
C03
VXLAN & EVPN Overlays
L2/L3 overlays with BGP-EVPN control plane, VNIs, and multi-tenant DC networking.
C04
Datacenter Storage Networking
iSCSI, NVMe-oF, FCoE and the storage fabrics that sit alongside compute.
C05
Server & Compute: CPU, GPU-Accelerators & NICs
Server internals, GPU accelerators, SmartNICs/DPUs and how compute drives fabric design.
C06
Lossless Ethernet — RoCEv2, PFC, ECN & DCQCN
The complete toolchain that makes commodity Ethernet behave as a lossless fabric for RDMA-based AI training traffic.
Module outcomes
- Compute PFC headroom buffers from link speed, cable length and MTU
- Design a Kmin/Kmax/Pmax ECN profile and justify the curve shape
- Trace an RP/NP/CP DCQCN interaction across multiple RTTs
- Design a complete PFC+ECN+DCQCN parameter set for a leaf-spine GPU fabric
C07
AI/ML Fabrics & GPU Cluster Networking
Connect hundreds-to-thousands of GPUs into a single training fabric: scale-up vs scale-out, rail-optimized Clos and collective-communication math.
Module outcomes
- Distinguish scale-up from scale-out interconnects and place each correctly in a GPU cluster
- Compare InfiniBand and Ethernet (RoCEv2/UEC) across cost, maturity and performance
- Design a rail-optimized Clos topology and compute non-blocking spine bandwidth for a GPU pod
- Calculate AllReduce completion time using Ring and Tree formulas
C08
Datacenter Automation & Fabric Management
Fabric-as-code with Apstra-style intent, Ansible and telemetry for DC operations.
C09
Datacenter Facilities: Power, Cooling & Cabling
Power/cooling budgets, structured cabling and the physical plant behind the fabric.
C10
Datacenter Design, Migration & Troubleshooting
End-to-end DC design, migration strategy and structured fabric troubleshooting.
The 7-asset system
Every module ships the same seven assets — 70 across this track.
You never get theory without a lab, a lab without commands, or a command without a way to verify it worked. Each of the 10 modules in Track C delivers all seven — closing with a 5-part workbook that hardens you for the exam and the job interview alike.
How the method worksStudy Guide
Focused conceptual grounding, scoped precisely to what the lab will exercise.
Lab Guide
Hands-on, step-by-step builds on real topologies — configure the actual technology, not a simulation.
Command Guide
A working reference of the exact CLI/API syntax used in the lab — muscle memory, not abstraction.
Glossary
Module-specific terminology that closes the language gap between generalist and specialist.
Solution Guide
Fully worked answers and reference configurations so learners self-correct and understand the reasoning.
Lab Outcomes & Verification
Explicit success criteria — the learner proves the lab works and produces evidence of competence.
5-Part Workbook
Exam-ready hardening: Foundations → Terminology → Concept Mastery → Design Challenges → Exam Simulation.
Career ladder
Where Track C takes you — with the salary band at each rung.
01Entry · 0-3 yrs
DC Network / Fabric Engineer
Rs 6-14 LPA
02Mid · 3-6 yrs
EVPN-VXLAN Fabric Engineer
Rs 14-26 LPA
03Senior · 6-10 yrs
AI Fabric / GPU-Cluster Architect
Rs 26-50 LPA
04Architect · 10+ yrs
Principal AI Infrastructure Architect
Rs 40-90+ LPA
Prove it with a certification
Track C maps directly to the DataCenter Stream.
The flagship ladder into India's AI-infrastructure gold rush — Clos fabrics, EVPN-VXLAN overlays, RoCEv2 lossless Ethernet and GPU-cluster design. Product-aligned to Cisco (Nexus, ACI) and Juniper (QFX, Apstra, JNCIA/JNCIP/JNCIE-DC).
Explore the DataCenter Stream10 modules. 70 assets. One track — deployable on day one.
The flagship track — spine-leaf fabrics, EVPN-VXLAN and lossless GPU-cluster networking.