Associate tierDataCenter streamLab-first · Rubric-graded

RCDARKR Certified DataCenter Associate

Build the fabric India's AI datacenters run on — from the first patch cable to a routed Clos underlay.

10 weeks8 hrs / week7 modules16 labs

Overview

What the RCDA certifies.

The RKR Certified DataCenter Associate (RCDA) is the entry point into the discipline every AI buildout in India ultimately depends on: the datacenter network. While generic IT roles are being compressed by automation, the physical and logical plumbing of AI infrastructure — racks, optics, spine-leaf fabrics, routed underlays — cannot be prompted into existence. RCDA takes a learner from zero datacenter exposure to independently cabling, configuring, and verifying a small spine-leaf pod on the two network operating systems that dominate Indian DC deployments: Cisco NX-OS (Nexus 9000) and Juniper Junos (QFX). Every module is lab-first — you earn the credential by doing the work on real virtual network operating systems (Nexus 9000v, vJunos-switch) in containerlab/EVE-NG pods, not by watching videos.

Technically, RCDA covers the modern DC canon at associate depth: facilities fundamentals (power, cooling, rack elevations, Uptime Institute tiers), structured cabling and optics (SMF/MMF, DAC/AOC, QSFP breakout), Layer 2 switching done right (802.1Q, MST, LACP, vPC), 3-stage Clos architecture with oversubscription math and ECMP, an eBGP underlay built to RFC 7938 conventions, and a first-principles introduction to VXLAN encapsulation — the on-ramp to the EVPN fabrics and lossless RoCEv2 GPU networks taught at the Professional tier. It is aligned in rigor and scope to Juniper JNCIA-DC and the datacenter foundations of Cisco CCNA, but it is RKR-owned, dual-vendor, and assessed by a graded practical exam — so the certificate is proof of demonstrated competence, not attendance.

Measurable outcomes

Walk out able to do this — on record.

Stand up a complete 2-spine/4-leaf Clos pod — cabling plan, L2 access, eBGP underlay, VXLAN segment — and prove it with CLI verification evidence

Operate confidently on both Cisco NX-OS and Juniper Junos, including the Junos candidate-config/commit model and NX-OS checkpoints/rollback

Select and validate physical media and optics (Cat6a, DAC, AOC, OM4/OS2 fiber, SFP+/QSFP28, 4x25G breakout) for a given rack row and distance budget

Design an access layer with 802.1Q trunks, MST, LACP port-channels, and Nexus vPC for dual-homed servers — and explain why L2 does not scale fabric-wide

Compute oversubscription ratios and size leaf uplinks for a stated east-west traffic profile

Configure and troubleshoot an eBGP IP-fabric underlay per RFC 7938 (private ASN plan, P2P /31s, ECMP multipath, BFD) on both NOSes

Explain and demonstrate VXLAN encapsulation end-to-end — VTEP, VNI, UDP 4789 — including a live packet capture and header dissection

Who it’s for

Built for these starting lines.

Final-year engineering students and fresh graduates targeting datacenter/NOC roles instead of shrinking generic IT support jobs

Enterprise campus/branch network engineers (CCNA-level) pivoting into datacenter and AI-infrastructure fabrics

DC facility technicians and smart-hands staff who want to move up the stack from cabling into switch configuration

IT support/system administrators who rack and connect servers and want to own the network side of the buildout

The syllabus

7 modules. 16 graded labs. No filler.

Every module terminates in a graded lab — theory is never left unproven. This is the full RCDA module sequence, exactly as delivered.

RCDA-M01

Datacenter Facilities & Physical Design

How a datacenter actually works below the network: Uptime Institute Tier I-IV availability topologies, power chain (utility feed, UPS, PDU, N/N+1/2N redundancy), cooling strategies (hot/cold aisle containment, in-row cooling, and why AI racks at 40-100+ kW break air cooling), rack units and elevations, and India's DC landscape (Mumbai/Chennai/Hyderabad hyperscale corridors). Establishes the vocabulary and constraints every fabric design lives inside.

You will be able to
  • Learner can classify a facility against Uptime Tier I-IV criteria and identify single points of failure in a power one-line diagram
  • Learner can compute rack power draw from equipment nameplate data and validate it against PDU and circuit capacity with headroom
  • Learner can produce a 42U rack elevation with correct airflow orientation, PDU placement, and cable entry planning
  • Learner can explain why 40 kW+ AI/GPU racks force liquid or rear-door cooling and what that means for network gear placement
Graded labs
Lab

Rack Elevation & Power Budget Design

Given a bill of materials (2 ToR switches, 16 dual-PSU servers, patch panels), build a 42U elevation, compute total draw vs dual 32A PDU capacity, place equipment for hot/cold aisle airflow, and defend the design against an injected failure (one PDU feed lost).

Uptime tier comparison chartRack elevation template (draw.io)Power budget calculator sheet
RCDA-M02

Structured Cabling, Optics & Transceivers

The physical layer as a first-class engineering discipline: copper (Cat6a, 10GBASE-T) vs DAC vs AOC trade-offs, multimode (OM3/OM4/OM5) and single-mode (OS2) fiber, transceiver families (SFP+, SFP28, QSFP+, QSFP28) and reach variants (SR/LR/DR/FR), 100G-to-4x25G breakout, MPO/MTP trunking, patch-panel discipline, and light-level verification from the switch CLI. Bad Layer 1 causes the majority of real DC incidents — this module makes learners the person who never causes them.

You will be able to
  • Learner can select the correct media and transceiver for a given distance, speed, and cost constraint (in-rack, cross-row, cross-hall)
  • Learner can read Tx/Rx light levels from NX-OS (show interface transceiver details) and Junos (show interfaces diagnostics optics) and judge them against spec thresholds
  • Learner can configure 100G-to-4x25G breakout on Nexus 9000 and QFX and explain the port-mapping consequences
  • Learner can document a patching schedule that a smart-hands technician can execute without ambiguity
Graded labs
Lab

Optics & Media Selection Workbench

Work through 12 real connectivity scenarios (leaf-to-spine cross-hall at 300 m, in-rack server attach, DCI at 8 km), pick media/optic for each, then validate two live links by reading DOM light levels on both NOSes and flagging one deliberately marginal link.

Lab

Breakout Cabling a Leaf Pod

Convert a 100G QSFP28 leaf port into 4x25G server-facing ports on Nexus 9000v and the QFX equivalent, bring up all four members, verify with show interface status / show interfaces terse, and produce the patching schedule for the row.

Optics reach cheat-sheet (SR/LR/DR/FR/ER)Fiber type + connector reference card
RCDA-M03

Layer 2 in the Datacenter: VLANs, MST, LACP & vPC

Classical DC access-layer switching done properly: 802.1Q VLANs and trunking, spanning-tree evolution (RSTP to MST) and why blocked links are wasted capacity, LACP port-channel formation and hashing, Nexus vPC (peer-link, peer-keepalive, consistency checks) for dual-homing servers, and the Junos view (aggregated Ethernet, MC-LAG/ESI-LAG awareness). Closes with the scaling argument — MAC table explosion, flood domains, STP convergence — that motivates routed Clos fabrics in M04.

You will be able to
  • Learner can configure VLANs and 802.1Q trunks consistently across NX-OS and Junos and verify tagging end-to-end
  • Learner can build LACP port-channels, explain hashing behavior for flow distribution, and diagnose a member that fails to bundle
  • Learner can deploy a Nexus vPC domain (peer-link, keepalive, role priority) and interpret vPC consistency-check failures
  • Learner can tune MST region configuration and predict which ports block for a given topology and priority set
  • Learner can articulate three concrete reasons large L2 domains fail at DC scale, with the failure mode for each
Graded labs
Lab

Dual-NOS VLAN & Trunking Bring-Up

Provision three tenant VLANs across a mixed Nexus 9000v + vJunos-switch access pair, trunk them, attach test hosts, and verify with show vlan / show ethernet-switching table plus end-to-end pings per VLAN.

Lab

vPC Dual-Homing a Server

Build a vPC domain on a Nexus pair (peer-keepalive over mgmt0, peer-link port-channel), dual-home a Linux host with an LACP bond, then break the peer-link and observe/document orphan-port and failover behavior.

Lab

MST Convergence Under Failure

In a looped L2 triangle, configure a single MST region with two instances, force root placement by priority, then fail the root port and measure reconvergence with timestamped pings — repeat with mismatched region names to see the failure mode.

vPC design checklistMST region planning worksheet
RCDA-M04

Spine-Leaf & Clos Fabric Architecture

Why every modern DC — and every GPU cluster — is a Clos network: 3-stage and 5-stage Clos topologies, leaf/spine/superspine roles, east-west vs north-south traffic, oversubscription ratio math (downlink vs uplink bandwidth), ECMP as the load-distribution engine, pod-based scaling, and failure-domain thinking. Learners compare a legacy core-aggregation-access design against a Clos redesign for the same server count and defend the difference quantitatively.

You will be able to
  • Learner can draw a 3-stage Clos for a stated port count and compute its oversubscription ratio at the leaf
  • Learner can explain why any leaf-to-leaf path in a Clos is exactly two hops and what that predictability buys AI/storage traffic
  • Learner can size spine count and uplink speeds for a given non-blocking or 3:1 target and justify the cost trade-off
  • Learner can map a Clos design onto physical reality: rack placement of leaves as ToR, spine placement, and cross-hall cabling volumes
Graded labs
Lab

Build the Pod: 2 Spines, 4 Leaves

Stand up the course's reference topology in containerlab (Nexus 9000v spines, mixed 9000v/vJunos leaves), produce the cabling matrix, assign P2P addressing from a /24 carved into /31s, and verify every link with LLDP neighbor checks.

Lab

Oversubscription Redesign Exercise

Given a traffic profile (48 servers per rack at 25G, measured 60% east-west utilization), compute the current 6:1 oversubscription, redesign uplinks to hit 3:1 and non-blocking targets, and present the port/optic bill-of-materials delta for each option.

Clos sizing calculatorReference pod topology file (containerlab YAML)
RCDA-M05

The Routed Underlay: eBGP IP Fabric (RFC 7938)

The associate-tier heart of the program: building the routed underlay that overlays ride on. Why eBGP beat OSPF/IS-IS as the hyperscale underlay choice (RFC 7938), private ASN numbering schemes (ASN-per-leaf, shared spine ASN), P2P /31 links, loopback advertisement, ECMP multipath (maximum-paths / multipath multiple-as), fast failure detection with BFD, and disciplined verification on both NX-OS and Junos. This is the module that separates RCDA holders from certificate collectors.

You will be able to
  • Learner can design an RFC 7938-style ASN and addressing plan (private 4-byte ASNs, /31 P2Ps, /32 loopbacks) for a pod and defend each choice
  • Learner can configure eBGP sessions leaf-to-spine on NX-OS and Junos and advertise loopbacks with correct policy/route-maps
  • Learner can enable and verify ECMP so all spine paths carry traffic, proving it with FIB output and flow tests
  • Learner can deploy BFD on fabric links and demonstrate sub-second failover versus BGP hold-timer defaults
  • Learner can work a structured troubleshooting flow for a session stuck in Idle/Active: interface, addressing, ASN mismatch, policy
Graded labs
Lab

eBGP Underlay Bring-Up (NX-OS)

On the reference pod's Nexus devices, implement the supplied ASN plan: eBGP on every fabric link, loopback advertisement via route-map, maximum-paths 4, then verify with show ip bgp summary, show ip route bgp, and traceroutes showing ECMP spread.

Lab

Junos Underlay & Multipath

Repeat the underlay on vJunos-switch leaves using policy-options and multipath multiple-as, exercising the candidate-config/commit workflow, and verify with show bgp summary and show route protocol bgp — then diff the operational models of the two NOSes in a short engineering note.

Lab

BFD & Failure Convergence Trial

Baseline convergence by killing a spine link with default BGP timers (measured with timestamped ping loss), then apply BFD (300ms x3) on both NOSes and re-measure — deliver a before/after convergence report.

RFC 7938 annotated summaryUnderlay verification runbook (both NOSes)
RCDA-M06

VXLAN Foundations: Overlay Networking Demystified

A rigorous first encounter with the overlay: why VXLAN exists (16M VNIs vs 4K VLANs, L2 adjacency over an L3 fabric), the encapsulation walk-through (original frame, VXLAN header, UDP 4789, outer IP between VTEP loopbacks), flood-and-learn with ingress replication, and hands-on static VXLAN on Nexus NVE interfaces and QFX. Ends with a clear-eyed look at flood-and-learn's limits — the setup for BGP EVPN at the RCDP Professional tier.

You will be able to
  • Learner can diagram the complete VXLAN encapsulation stack and state the size and purpose of each header field that matters operationally
  • Learner can configure a static VXLAN segment (VLAN-to-VNI mapping, NVE/VTEP source loopback, ingress-replication peer list) on Nexus 9000v
  • Learner can verify overlay health with show nve peers, show nve vni, and MAC learning tables, and correlate them to underlay reachability
  • Learner can capture live VXLAN traffic and identify VNI, VTEP addresses, and inner frame in a Wireshark dissection
  • Learner can explain two scaling problems of flood-and-learn that EVPN's control plane solves
Graded labs
Lab

Static VXLAN Segment Bring-Up

Map VLAN 10 to VNI 10010 across two Nexus 9000v leaves over the M05 underlay: configure NVE1 with source-interface loopback1 and ingress-replication protocol static, attach hosts, and prove L2 adjacency across the routed fabric with pings plus show nve peers.

Lab

VXLAN Packet Autopsy

Run tcpdump on a fabric link while overlay hosts communicate, open the capture in Wireshark, and annotate a single packet end-to-end: outer MACs/IPs (VTEP loopbacks), UDP source-port entropy for ECMP, VNI field, and the untouched inner frame.

VXLAN header anatomy posterAnnotated capture file (pcap)
RCDA-M07

Operating the Fabric: NX-OS & Junos Ops Discipline

The operational craft that makes an associate immediately employable: NX-OS configuration management (checkpoints, rollback, config sessions) versus the Junos commit model (candidate config, commit confirmed, rollback 0-49, commit comments), software image management and upgrade hygiene, out-of-band management networks, syslog and SNMPv3 setup, NTP discipline, and a repeatable interface-down troubleshooting method. Closes with a graded incident-ticket simulation run against the full pod built across the course.

You will be able to
  • Learner can use Junos commit confirmed and rollback to make risk-free changes, and replicate the discipline on NX-OS with checkpoints
  • Learner can configure syslog, SNMPv3, and NTP on both NOSes so the pod is observable and time-consistent
  • Learner can execute a structured L1-to-L3 diagnosis of a down or degraded link (optics DOM, errors/CRC, LLDP, protocol state) in under 15 minutes
  • Learner can write an incident ticket update that states symptom, evidence, root cause, fix, and verification — the format used in real NOC handoffs
Graded labs
Lab

Commit & Rollback Fire Drill

On vJunos-switch, push a deliberately breaking change under commit confirmed 3 and watch auto-rollback save the session; then replicate the safety net on NX-OS with checkpoint/rollback running-config checkpoint — documenting the exact command sequences for both.

Lab

Observability Baseline

Configure the full pod to send syslog to a collector VM, enable SNMPv3 with authPriv, sync NTP, and prove it by generating a link-flap event and locating it in the collector within 60 seconds.

Lab

Graded Incident Simulation

The instructor injects one of six faults (wrong VNI mapping, BFD misconfig, ASN typo, marginal optic, MTU mismatch, vPC inconsistency) into your pod; you have 45 minutes to diagnose, fix, verify, and file a NOC-quality ticket — scored on the RCDA practical rubric.

Dual-NOS command translation cardIncident ticket templateTroubleshooting decision tree

How you’re examined

The RCDA exam format.

Two-part RKR assessment, both proctored. Part 1 — Theory (90 minutes, online proctored): 60 scenario-based questions across the seven modules, including exhibit questions with real NX-OS/Junos CLI output and cabling/optics matching items; pass mark 70%. Part 2 — Practical Lab Exam (3 hours, remote-proctored on RKR's cloud lab): the candidate receives a broken/partially built 2-spine, 4-leaf pod (Nexus 9000v + vJunos-switch) and must complete graded tasks — fix a mis-cabled breakout port, bring up LACP/vPC to a dual-homed host, configure the eBGP underlay per a supplied RFC 7938 ASN plan, map a VLAN to a VXLAN VNI with ingress replication, and produce verification output (show bgp summary, show nve peers, show interfaces diagnostics optics). Tasks are auto-graded against device state plus a rubric-scored verification report; lab pass mark 75%. Both parts must be passed within a 90-day window; one free lab retake included.

Career plan

Where the RCDA takes you.

RCDA is engineered to land the first datacenter role and compress the entry-to-mid climb. India's DC buildout is chronically short of people who can be trusted alone in a pod — RCDA's graded practical exam is designed as exactly that proof. Holders enter DC NOC and deployment roles, and the dual-NOS + eBGP underlay skill set is the direct feeder into fabric engineering, where the RCDP Professional tier (EVPN, multitenancy, lossless RoCEv2 for GPU clusters) unlocks senior bands.

Roles unlocked
Datacenter NOC Engineer (L1/L2)Datacenter Network Engineer (deployment/operations)Network Field/Implementation Engineer — DC buildoutsJunior Fabric Engineer (underlay operations)DC Infrastructure Technician (network-side, colocation/hyperscale)
Salary band
Rs 6-14 LPA entry band, with a credible 14-20 LPA step within 2-3 years on the fabric track
Rung 1 · 0-1 yr
DC NOC Engineer / Network Technician
Rs 6-9 LPA
Rung 2 · 1-3 yrs
Datacenter Network Engineer (deploy & operate pods)
Rs 9-14 LPA
Rung 3 · 3-5 yrs
Fabric Engineer — EVPN/VXLAN operations (RCDP track)
Rs 14-20 LPA
Rung 4 · 5+ yrs
Senior DC Network Engineer / GPU fabric specialist
Rs 20-26 LPA
Demand signal

As of mid-2026, India's datacenter capacity is scaling from roughly 1,700 MW toward 5-6.5 GW by 2030 — a buildout forecast to create ~100,000 datacenter jobs — yet 73% of DC operations roles are reported hard to fill, niche network-infrastructure skills command a ~1.7x pay premium, and India faces a ~53% AI-skills gap in 2026.

Continue the ladder

More certifications in the DataCenter Stream.

Professional

RCDP — RKR Certified DataCenter Professional

Own the fabric: EVPN-VXLAN, ACI, and Apstra — with the lossless-Ethernet on-ramp to AI datacenters.

16 weeks 18 labs₹99,999
Expert

RCDE — RKR Certified DataCenter Expert

Design the fabrics that train India's AI — and defend every decision under fire.

20 weeks 19 labs₹1,99,999

7 modules. 16 graded labs. One verifiable credential.

10 weeks at 8 hours a week — proven at the lab pod, scored against a published rubric.

Compare all certifications