Professional tierDataCenter streamLab-first · Rubric-graded

RCDPRKR Certified DataCenter Professional

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

16 weeks10 hrs / week8 modules18 labsPrerequisite: RCDA

Overview

What the RCDP certifies.

The RKR Certified DataCenter Professional (RCDP) is the fabric-ownership tier of RKR's datacenter track. Where the associate level proved you can configure switches, RCDP proves you can design, deploy, and operate an entire multi-tenant datacenter fabric — the EVPN-VXLAN overlay architecture that underpins virtually every serious Indian datacenter built since 2020, plus the two controller philosophies competing to manage it: Cisco ACI's application-centric policy model and Juniper Apstra's intent-based blueprints. The program is deliberately dual-stack — every core concept is exercised on both Nexus 9000v/NX-OS and vJunos/Junos — because India's colo and enterprise DC market is mixed-vendor and the engineers who move fluidly between them command the premium.

RCDP is also the deliberate on-ramp to AI infrastructure. India's datacenter capacity is scaling from roughly 1,700 MW toward 5-6.5 GW by 2030, and an increasing share of that buildout is GPU and storage fabric that cannot tolerate packet loss. The storage networking module (NVMe/TCP, NVMe/RoCE) and the lossless Ethernet module (PFC, ECN, DCQCN) give RCDP graduates the exact vocabulary and hands-on evidence that AI-fabric hiring panels screen for — while the automation spine (fabric-as-code, REST-driven ACI, gNMI telemetry, CI-gated change) reflects RKR's core thesis: the roles being automated away are the ones typing configs by hand; the roles being created are the ones building the automation. Eighteen graded labs and an 8-hour practical exam mean every line on the certificate is demonstrable.

Measurable outcomes

Walk out able to do this — on record.

Design and deploy production-grade EVPN-VXLAN fabrics — Clos underlay, symmetric IRB, anycast gateways, multi-tenancy, and Type-5 DCI — on both NX-OS and Junos platforms.

Operate controller-led datacenters two ways: the Cisco ACI policy model (EPGs, contracts, L3Outs, VMM) and Juniper Apstra intent-based blueprints with IBA-driven day-2 operations.

Engineer the network for storage: NVMe/TCP and NVMe/RoCEv2 transport design, dedicated storage segments, and honest fio-benchmarked validation.

Configure and defend lossless Ethernet foundations — PFC no-drop classes, ECN/WRED, DCQCN behavior — the entry skill for AI/GPU fabric roles.

Deliver every fabric change as code: YAML source of truth, Jinja2/Ansible deployment, REST-driven ACI, CI-gated validation, and gNMI streaming telemetry.

Troubleshoot under pressure: pass a ten-fault gauntlet spanning underlay, overlay, policy, and QoS with documented root cause and time-to-restore.

Who it’s for

Built for these starting lines.

RCDA graduates ready to move from device configuration to fabric-level design and ownership

Mid-level enterprise or colo network engineers (2-5 years) whose sites are migrating to EVPN-VXLAN, ACI, or Apstra

Virtualization and storage engineers extending into the network side of NVMe-oF and converged fabrics

NOC and operations engineers targeting senior fabric-operations roles at hyperscaler-adjacent and GPU-cloud facilities

Presales and delivery engineers who must design and defend multi-tenant fabric proposals with technical credibility

The syllabus

8 modules. 18 graded labs. No filler.

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

RCDP-M01

EVPN-VXLAN Fabric Foundations

Build the modern datacenter fabric from the ground up: a 3-stage Clos IP underlay using eBGP unnumbered (RFC 5549-style) with ECMP and BFD, then a VXLAN data plane driven by an MP-BGP EVPN control plane. You will work hands-on with EVPN route types 1 through 5 on both Cisco Nexus 9000v (NX-OS) and vJunos-switch, comparing ingress replication against multicast underlay for BUM traffic and dissecting real VXLAN packets on the wire.

You will be able to
  • Learner can design and configure an eBGP-unnumbered Clos underlay with ECMP and BFD across Nexus 9000v and vJunos-switch leaves and spines.
  • Learner can explain EVPN route types 1-5 and verify Type-2 MAC/IP and Type-3 IMET advertisement with platform show commands on both NX-OS and Junos.
  • Learner can configure VXLAN VTEPs with ingress replication and validate VNI-to-VLAN mapping and NVE peering end to end.
  • Learner can troubleshoot BUM forwarding, MAC mobility (sequence-number races), and mass-withdraw behavior in a live fabric.
  • Learner can justify replication-mode and underlay-protocol choices for a given fabric scale in a written design note.
Graded labs
Lab

Clos Underlay Build

Bring up a 2-spine / 4-leaf underlay with eBGP unnumbered, ECMP, and BFD on a mixed Nexus 9000v + vJunos pod; verify ECMP hashing and sub-second link-failure convergence.

Lab

EVPN Control-Plane Dissection

Layer L2 EVPN-VXLAN on the underlay, capture Type-2/Type-3 routes from the BGP table, and decode VXLAN encapsulation (VNI, UDP 4789) in Wireshark from a SPAN capture.

Mixed-vendor virtual pod (Nexus 9000v + vJunos-switch)EVPN route-type reference cardWireshark capture set with annotated VXLAN flows
RCDP-M02

Multi-Tenant Overlay Design & DCI

Turn a flat overlay into a real multi-tenant service platform. This module covers symmetric IRB with per-VRF L3VNIs, distributed anycast gateways, and RD/RT numbering schemes that survive multi-pod growth, then extends tenants across sites with EVPN Type-5 DCI through border leaves. Controlled route leaking for shared services and hard isolation proofs are graded, not assumed.

You will be able to
  • Learner can implement multi-tenant L3 services using symmetric IRB, per-VRF L3VNIs, and distributed anycast gateways.
  • Learner can design an RD/RT and VNI numbering scheme that scales across pods and provably prevents cross-tenant leakage.
  • Learner can configure controlled inter-VRF route leaking to expose shared services (DNS, NTP, monitoring) without collapsing isolation.
  • Learner can stretch tenants across two sites using EVPN Type-5 routes on border leaves and measure failover behavior under DCI link loss.
Graded labs
Lab

Three-Tenant Fabric

Build three isolated tenants with anycast gateways on a shared fabric; demonstrate isolation with traffic tests, then surgically leak a shared-services VRF and re-verify boundaries.

Lab

Type-5 DCI Stretch

Interconnect two independent EVPN fabrics via border-leaf Type-5 routes; fail the primary DCI path and record convergence, route churn, and traffic loss.

Multi-pod addressing and RD/RT design workbookDCI failover test plan template
RCDP-M03

Cisco ACI Policy Model & Operations

Master the application-centric alternative to box-by-box configuration: APIC-driven fabrics where tenants, application profiles, EPGs, bridge domains, and contracts define connectivity as policy. You will build L3Outs with OSPF and BGP peering to the outside world, integrate a VMM domain with vCenter, and learn to debug the fabric the way TAC does — through health scores, the fault lifecycle, moquery, and the REST object model.

You will be able to
  • Learner can model a multi-tier application in ACI using tenants, application profiles, EPGs, bridge domains, and contracts with filters.
  • Learner can configure an L3Out with OSPF or BGP peering to external routers and control external EPG classification and route export.
  • Learner can integrate a VMM domain with vCenter and validate dynamic EPG-to-port-group binding for virtualized workloads.
  • Learner can troubleshoot ACI using health scores, the fault lifecycle, moquery, and the API Inspector to trace policy resolution to concrete zoning rules.
  • Learner can push tenant configuration through the APIC REST API and verify it against the managed-object tree.
Graded labs
Lab

App-Centric Build

On the ACI simulator, model a 3-tier application: web/app/db EPGs, bridge domains, and contracts; prove that traffic flows only along contract-permitted paths.

Lab

L3Out & External Connectivity

Configure an L3Out with BGP to an external Nexus router, classify external prefixes into an external EPG, and control advertisement of tenant subnets outward.

Lab

Object-Model Debugging

Use moquery and the API Inspector to trace a broken contract to its zoning-rule entries, then repair and redeploy the tenant via a REST POST from Python.

ACI simulator accessContract and zoning-rule troubleshooting flowchartPostman/Python REST collection for tenant provisioning
RCDP-M04

Juniper Apstra Intent-Based Fabrics

Operate the fabric by intent instead of by CLI: Apstra device profiles, logical devices, rack types, and templates compile into a deployable EVPN-VXLAN blueprint whose rendered configuration you can inspect line by line. The module goes deep on day-2 reality — routing zones, connectivity templates, intent-based analytics (IBA) probes, drift detection, and staged changes with full rollback.

You will be able to
  • Learner can compose an Apstra blueprint from device profiles, rack types, and templates and deploy a working EVPN-VXLAN fabric to vJunos devices.
  • Learner can create routing zones (tenant VRFs), virtual networks, and connectivity templates for server-facing interfaces.
  • Learner can configure IBA probes to surface ECMP imbalance, MTU mismatch, and cabling drift, and interpret raised anomalies.
  • Learner can execute day-2 changes — adding a leaf, expanding a virtual network — through staged blueprints with commit and rollback.
Graded labs
Lab

Intent to Fabric

Model rack types and a 3-stage template, deploy the blueprint to a vJunos pod, and diff Apstra's rendered configuration against hand-built configs from M01.

Lab

Drift & Anomaly Hunt

An instructor harness injects out-of-band CLI drift and a miscabled leaf; detect both through Apstra anomalies and IBA probes, then remediate from the controller.

Apstra lab instance with pre-registered device profilesIBA probe cookbook
RCDP-M05

Storage Networking & NVMe over Fabrics

Datacenter fabrics increasingly carry the storage, not just the app traffic. This module covers the NVMe-oF family — NVMe/TCP and NVMe/RoCEv2 — against the iSCSI and Fibre Channel incumbents: queue models, discovery controllers, ANA multipathing, and what each transport demands from the network. You will design dedicated storage segments (VRF separation, MTU 9216, QoS marking) and benchmark them honestly with fio.

You will be able to
  • Learner can compare NVMe/TCP, NVMe/RoCEv2, iSCSI, and FC on latency, loss tolerance, and fabric requirements to justify a transport choice for a given workload.
  • Learner can configure a Linux NVMe/TCP target and initiator across a routed EVPN fabric and baseline IOPS and latency with fio.
  • Learner can design a storage-dedicated fabric segment with VRF isolation, jumbo MTU, and DSCP marking for host-to-storage traffic.
  • Learner can troubleshoot NVMe-oF discovery failures, ANA multipathing state, and MTU blackholes end to end.
Graded labs
Lab

NVMe/TCP End-to-End

Stand up a Linux NVMe/TCP target and initiator across the EVPN fabric, record a fio baseline, then break path MTU mid-run and diagnose the blackhole from symptoms to fix.

Lab

Storage Segment Design

Carve a dedicated storage VRF with jumbo MTU and QoS marking; benchmark NVMe/TCP on the protected path versus a congested best-effort path and defend the delta.

Linux storage node images (nvmet target + nvme-cli)fio benchmark methodology sheet
RCDP-M06

Lossless Ethernet & RoCEv2 Introduction

The on-ramp to AI and high-performance storage fabrics: why RDMA traffic over RoCEv2 (UDP 4791) cannot tolerate drops, and how PFC no-drop classes, buffer headroom, ECN/WRED marking, and DCQCN congestion control combine to deliver lossless behavior on ordinary Ethernet. Equal weight is given to the failure modes — PFC storms, head-of-line blocking, and the watchdog and monitoring discipline that keeps them contained.

You will be able to
  • Learner can explain the RoCEv2 packet flow and why RDMA workloads (GPU collectives, NVMe/RoCE) require lossless or ECN-managed transport.
  • Learner can configure PFC on a dedicated no-drop class with correct buffer headroom and validate pause-frame behavior under load.
  • Learner can enable ECN/WRED thresholds and observe DCQCN CNP generation and sender rate reduction during congestion.
  • Learner can detect and mitigate PFC storms and head-of-line blocking using PFC watchdog and per-queue counters.
Graded labs
Lab

Build a No-Drop Class

Classify RoCEv2 traffic (DSCP 26 / UDP 4791) into a PFC-enabled queue on Nexus and SONiC images; drive incast congestion with a traffic generator and prove zero drops in the no-drop class.

Lab

ECN in Action

Tune WRED/ECN min-max thresholds, replay an incast pattern, and graph CNP counts, queue depth, and throughput recovery to show DCQCN reacting in real time.

SONiC virtual switch imagesIncast traffic-generation scriptsPFC/ECN tuning worksheet
RCDP-M07

Fabric Automation & Streaming Telemetry

The professional fabric is deployed from a source of truth, not typed at a prompt. Build a fabric-as-code pipeline: YAML intent rendered through Jinja2 and deployed with Ansible to NX-OS and Junos, ACI provisioned idempotently over REST from Python, and every change gated by pre/post validation with Nornir/NAPALM or pyATS in CI. Close the loop with gNMI/OpenConfig streaming telemetry feeding Grafana dashboards and alerts.

You will be able to
  • Learner can generate complete leaf/spine EVPN configurations from a YAML source of truth using Jinja2 templates and deploy them with Ansible to NX-OS and Junos.
  • Learner can provision ACI tenants and EPGs idempotently from Python via the REST API, with rollback on failure.
  • Learner can implement pre/post-change validation (BGP session state, VNI counts, route-table diffs) with Nornir/NAPALM or pyATS inside a CI pipeline.
  • Learner can stream gNMI/OpenConfig telemetry into a Telegraf-InfluxDB-Grafana stack and build alerts on fabric-health KPIs.
Graded labs
Lab

Fabric-as-Code

Model the M01 fabric as YAML, render with Jinja2, deploy with Ansible; a single git commit adding a leaf triggers the pipeline to render, validate, and deploy it.

Lab

ACI via API

Write a Python provisioning script that creates a tenant, EPGs, and contracts through the APIC REST API, is safe to run twice, and rolls back cleanly on a simulated mid-run failure.

Lab

Telemetry Pipeline

Subscribe to gNMI interface and BGP paths, land the streams in InfluxDB via Telegraf, and build a Grafana dashboard with alerts that fire during an induced link flap.

Reference fabric-as-code repositoryCI pipeline templates (GitLab)TIG stack compose file
RCDP-M08

Professional Capstone: Multi-Tenant Fabric Build & Defense

A timed, graded rehearsal of the RCDP practical exam. From a written requirements document, design and build a greenfield fabric — underlay, two EVPN tenants with symmetric IRB and Type-5 DCI, an ACI application segment, a PFC/ECN storage class, and one change delivered through the automation pipeline. Then survive the fault gauntlet: ten injected failures spanning every layer, scored on time-to-restore and the quality of your root-cause notes, followed by a design defense.

You will be able to
  • Learner can translate a written multi-tenant requirements document into a validated design artifact (IP plan, RD/RT scheme, VNI allocation) and implement it within a time budget.
  • Learner can localize and repair injected faults spanning underlay routing, EVPN control plane, ACI policy, and QoS within SLA-style time limits.
  • Learner can produce as-built documentation and defend design trade-offs — replication mode, IRB model, DCI method — in a live review with an RKR examiner.
Graded labs
Lab

Greenfield Build-Off

An 8-hour graded build from a requirements document on a fresh virtual pod, scored by the same automated validation harness used in the RCDP-PX practical.

Lab

Fault Gauntlet

Ten injected faults across underlay, overlay, ACI, storage, and QoS; scored on time-to-restore, blast-radius containment, and written root-cause analysis.

Exam-style requirements documents (three variants)Automated grading harnessDesign-defense rubric

How you’re examined

The RCDP exam format.

Two-part RCDP-PX assessment. Part 1: proctored theory — 85 scenario-based questions (topology exhibits, CLI output analysis, design trade-off items) in 120 minutes, pass mark 72%. Part 2: an 8-hour graded practical on a live virtual pod running Cisco Nexus 9000v, vJunos-switch, the ACI simulator, and Juniper Apstra — candidates build a two-tenant EVPN-VXLAN fabric with symmetric IRB and Type-5 DCI from a written requirements document, deploy an ACI application profile with contracts and an L3Out, configure a PFC/ECN no-drop class for RoCEv2 traffic, and execute one change through the fabric-as-code pipeline; a fault-injection segment scores time-to-restore and written root-cause notes. Graded against a published rubric by RKR examiners plus an automated validation harness; overall pass at 70% with minimum bars in every domain. One free retake of the practical within 90 days.

Career plan

Where the RCDP takes you.

RCDP moves engineers from device-level datacenter work into fabric-level ownership — the mid-to-senior band where India's DC buildout is bidding hardest. EVPN-VXLAN, ACI, and Apstra skills sit on nearly every hyperscaler, colo, and GPU-cloud job description written in 2026, and the lossless-Ethernet module positions graduates one step from AI-fabric roles that carry the steepest premiums. The graded capstone and fault gauntlet give hiring managers verifiable evidence, not a watch-time certificate.

Roles unlocked
Senior DataCenter Network Engineer (EVPN-VXLAN fabrics)ACI Fabric Engineer / ACI Operations SpecialistNetwork Automation Engineer — DC FabricDataCenter Fabric Lead (colo / GPU-cloud / hyperscaler-adjacent)Intent-Based Networking Engineer (Apstra / controller-led operations)
Salary band
Rs 14-26 LPA on certification, with a senior-track runway to Rs 26-50 LPA
Entry point (typical RCDA holder)
DataCenter Network Engineer
Rs 8-14 LPA
0-12 months post-RCDP
Senior DC Fabric Engineer — EVPN-VXLAN / ACI
Rs 14-26 LPA
1-3 years
Lead Fabric Engineer / Automation-First Fabric Owner
Rs 26-40 LPA
3-5 years (RCDE / expert track)
Principal DataCenter Fabric Architect
Rs 40-50 LPA (Rs 50-90+ on the expert/AI-fabric track)
Demand signal

June 2026: India's live datacenter capacity has crossed ~1,700 MW and is tracking to 5-6.5 GW by 2030 (~100,000 DC jobs), yet 73% of operators report fabric and operations roles hard to fill — and engineers with EVPN, ACI, and lossless-Ethernet depth are commanding a 1.7x premium over generalist network pay.

8 modules. 18 graded labs. One verifiable credential.

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

Compare all certifications