OcNOS-DC  ·  EVPN-VXLAN  ·  BGP Leaf-Spine  ·  Open Hardware  ·  24/7 SLA

DC Fabric

Scale-out leaf-spine data center fabric with EVPN-VXLAN overlays. Multi-tenant workload isolation. BGP ECMP at every tier — on validated open hardware at a fraction of proprietary cost.

Protocols:
EVPN-VXLAN BGP Wide ECMP VXLAN PFC / DCB BFD OpenConfig gNMI gRPC Streaming Telemetry NETCONF / YANG Advanced Load Balancing Multisite Stitching IP Storage / NVMEoF 3rd-Party Optics Model-Based OS
Non-Blocking
Wide-ECMP fabric architectures
EVPN
RFC 7432 / 8365 · Multi-Homing
3–5×
TCO savings vs proprietary switches
24/7
Carrier-grade global SLA
The Problem

Proprietary switches own your hardware refresh cycle forever.

Every time you buy a Cisco or Arista switch, you're paying a hardware margin and a software license to the same vendor — with no ability to separate them. OcNOS-DC runs on the same Broadcom ASICs at ODM prices.

EVPN-VXLAN leaf-spine with BGP underlay anycast gateway ARP suppression EVPN multi-homing — production-ready, on Day 1, on open hardware.

Reference Architecture

EVPN-VXLAN Leaf-Spine — RFC 7938 + RFC 8365

eBGP unnumbered underlay on every leaf-spine link. VXLAN VTEP on every leaf. Anycast gateway eliminates first-hop routing latency. ZTP provisions every new switch at boot.

EVPN-VXLAN leaf-spine DC fabric with border leaf Two-spine, two-leaf EVPN-VXLAN data center fabric. Tenant A and Tenant B servers attach via ESI-LAG to VTEP leaves running OcNOS-DC with Type-2 and Type-5 EVPN routes and anycast gateway. Two spines reflect EVPN routes over eBGP ECMP at 25.6 Tbps each. A border leaf with 400G uplinks provides external WAN connectivity and EVPN gateway. Servers Tenant A ESI-LAG AA Servers Tenant B ESI-LAG AA Servers Tenant A anycast GW Leaf-01 OcNOS-DC VTEP · anycast GW Type-2 + Type-5 eBGP AS 65001 EVPN-VXLAN VTEP Leaf-02 OcNOS-DC eBGP AS 65002 EVPN-VXLAN VTEP eBGP ECMP Spine-01 OcNOS-DC 25.6 Tbps · RR EVPN Route Reflector eBGP AS 65100 Spine-02 OcNOS-DC 25.6 Tbps · RR EVPN Route Reflector eBGP AS 65101 Border Leaf OcNOS-DC Type-5 · RPKI 400G uplinks EVPN GATEWAY WAN / inet OcNOS-DC — DC FABRIC — EVPN-VXLAN · BGP LEAF-SPINE · ANYCAST GW · MULTI-TENANT
OcNOS-DC leaf (VTEP)
OcNOS-DC spine (RR)
OcNOS-DC border leaf
External / WAN

Hover nodes for capability and platform details  ·  Full HCL: 40+ validated platforms at ipinfusion.com/hcl

600+Operator Deployments
60+Countries
26Years in Networking
Technical Architecture

EVPN-VXLAN done right — and automated from Day 0.

Most data center fabric deployments still rely on manual CLI provisioning that takes days per rack. OcNOS-DC combines the full EVPN-VXLAN feature set with ZTP, gNMI streaming telemetry, and Ansible automation — so you go from bare switch to production leaf in minutes.

EVPN Type-2 + Type-5 with Symmetric IRB

Type-2 routes advertise MAC and IP together so every leaf knows every host without flooding. Type-5 routes distribute IP prefixes for inter-tenant and external routing. Symmetric IRB gives you L3 forwarding on both ingress and egress VTEPs — optimal path, no hairpin.

EVPN Multi-Homing — active-active, no MLAG

Servers dual-homed to two leaf switches using ESI-LAG. Both links forward traffic simultaneously, with EVPN handling MAC/IP synchronization across VTEPs. No proprietary MLAG domain, no peer-link, no split-brain risk.

Distributed Anycast Gateway + ARP Suppression

Every leaf switch shares the same gateway IP and MAC for each tenant subnet — no traffic trombones through a central gateway. ARP suppression means leaf switches answer ARP locally from the EVPN database instead of flooding, reducing broadcast by 80–90% at scale.

ZTP + gNMI — zero-touch from rack to production

A new OcNOS-DC switch boots, DHCP provisions its full EVPN config via ZTP, and gNMI on-change telemetry begins streaming within minutes. No console cable, no manual eBGP configuration. Ansible and Terraform handle Day-2 changes idempotently across the entire fabric.

ansible — deploy EVPN-VXLAN fabric LIVE
$ansible-playbook -i dc-inventory deploy-evpn.yml
ok: [spine-01] — EVPN RR, AS 65100
ok: [spine-02] — EVPN RR, AS 65101
ok: [leaf-01] — VTEP 10.0.0.1, VNIs up
ok: [leaf-02] — VTEP 10.0.0.2, ESI peer
ok: [border-01] — Type-5, RPKI active
PLAY RECAP
5 devices changed=5 failed=0 unreachable=0
$gnmic sub --path /network-instances/ \
--mode ON_CHANGE
VNI 10010 up MACs: 248 ARPs: 248
VNI 10020 up MACs: 192 ARPs: 192
VNI 50010 up routes: 14
✓ fabric healthy — 0 drops — ECMP balanced

Validated DC Fabric Platforms

AS9736-64D
Edgecore
25.6 Tbps
AS7726-32X
Edgecore
6.4 Tbps
S9600-56DX
UfiSpace
4.8 Tbps
AS7726-32X
Celestica
6.4 Tbps

40+ validated platforms — view full HCL →

Open vs Proprietary

The same ASICs. A fraction of the cost.

Arista, Cisco, and Juniper data center switches run Broadcom Tomahawk and Trident ASICs — the same silicon inside Edgecore and UfiSpace ODM hardware. What you're paying extra for is the brand, the lock-in, and the vendor's margin.

❌  Proprietary (Arista / Cisco / Juniper)

Hardware and NOS bundled — you cannot buy the ASIC without the software license attached to it.

MLAG for server dual-homing requires a proprietary peer-link protocol and dedicated keepalive link per switch pair.

Automation uses vendor-specific EAPI, NX-API, or Junos PyEZ — not portable across platforms or vendors.

Separate hardware support and software support contracts — multiple TAC queues for the same network problem.

Forklift refresh every 3–5 years when the vendor EOLs the platform, whether or not the hardware has failed.

✓  OcNOS-DC on Open Hardware

Hardware from Edgecore or UfiSpace. OcNOS-DC from IP Infusion. Separate contracts — or bundled via a single Turnkey SKU.

EVPN Multi-Homing (ESI-LAG) delivers active-active server dual-homing natively — no peer-link, no proprietary protocol.

NETCONF/YANG 1.1, OpenConfig, gNMI, Ansible, Terraform — open standards that work on OcNOS-DC and every other standards-compliant NOS.

Single IP Infusion support contract covers software TAC and hardware RMA coordination globally, 24/7 — one call, one queue.

OcNOS-DC is software — upgrade the NOS independently of the hardware. Extend platform life beyond vendor EOL timelines.

Deployment Scenarios

Where OcNOS DC Fabric is deployed today.

USE CASE 01

Enterprise Multi-Tenant DC Fabric

Enterprise data centers hosting multiple business units or customers on shared infrastructure need strong tenant isolation without sacrificing east-west performance. OcNOS-DC delivers per-VRF isolation via L3 VNIs, anycast gateway for optimal intra-fabric routing, and ARP suppression to contain broadcast at scale — all on open hardware.

USE CASE 02

Cloud Provider / Hyperscale Fabric

Cloud providers and hyperscalers need automation-first operations where every switch is provisioned identically from a template, never manually. OcNOS-DC's ZTP, NETCONF/YANG, and Ansible-native design means new capacity is operational in minutes, not days. Docker and Kubernetes support in OcNOS 7.0 lets operators run on-switch monitoring agents alongside the NOS.

USE CASE 03

Mixed AI + General-Purpose Fabric

Many data centers run both GPU compute pods and conventional workloads. OcNOS-DC handles both in a single NOS instance — EVPN-VXLAN multi-tenant fabric for the conventional workloads and the full lossless RoCEv2 DCB stack (PFC, ECN, DLB) for GPU pods. No separate NOS for the AI section. For the DC-to-DC interconnect layer, see the IPoDWDM solution.

Get Started

Build your leaf-spine fabric on open hardware.

Talk to an IPI DC networking specialist, or download the free OcNOS-DC VM and validate EVPN-VXLAN configurations today — no hardware required.

Common Technical Questions

DC Fabric with OcNOS-DC

What is EVPN-VXLAN and how does OcNOS-DC implement it?
EVPN-VXLAN is the industry standard for multi-tenant data center fabric overlays. OcNOS-DC implements RFC 8365 EVPN with VXLAN encapsulation: Type-2 routes for MAC/IP advertisement, Type-5 routes for IP prefix distribution, and both symmetric and asymmetric Integrated Routing and Bridging (IRB). Leaf switches act as VXLAN Tunnel Endpoints (VTEPs) with distributed anycast gateway for optimal intra-fabric routing. ARP suppression reduces broadcast traffic at scale. The BGP underlay follows RFC 7938 data center architecture with eBGP unnumbered interfaces on every leaf-spine link.
Can OcNOS-DC replace Arista EOS or Cisco NX-OS in my data center?
Yes. OcNOS-DC provides feature parity for standard EVPN-VXLAN leaf-spine deployments. It runs on the same Broadcom Tomahawk and Trident ASICs used in Arista and Cisco hardware, but on open ODM platforms from Edgecore, UfiSpace, and Celestica. Customers typically achieve substantial CAPEX savings on hardware versus proprietary systems. IP Infusion provides a single unified support contract covering both OcNOS-DC software and hardware RMA coordination — equivalent operational simplicity to a proprietary deployment, without the vendor lock-in.
What is EVPN Multi-Homing and does OcNOS-DC support it?
EVPN Multi-Homing (ESI-LAG) allows a server or network device to connect to two or more leaf switches simultaneously in an active-active or active-standby configuration, without requiring a traditional MLAG domain. OcNOS-DC supports EVPN Multi-Homing (Active-Active) using Ethernet Segment Identifiers (ESI), providing redundancy and optimal traffic distribution without the operational complexity of proprietary MLAG protocols.
How does OcNOS-DC handle multi-tenant isolation in a DC fabric?
Multi-tenant isolation is implemented via VXLAN Network Identifiers (VNIs). OcNOS-DC supports both Layer 2 VNIs for bridged tenant domains and Layer 3 VNIs for routed inter-tenant traffic via symmetric IRB. Each tenant VRF is isolated at the data plane. EVPN E-Tree (Scenario 1) provides additional leaf-to-leaf isolation within a tenant for use cases requiring spoke-to-spoke traffic control. BGP RPKI validation secures the routing control plane at the fabric edge.
Does OcNOS-DC support Zero Touch Provisioning for leaf-spine deployments?
Yes. OcNOS-DC supports DHCP-based ZTP over both IPv4 and IPv6. A new leaf switch boots, requests its configuration via DHCP option 67, and OcNOS-DC downloads and applies the full configuration automatically — including eBGP sessions, EVPN address families, VNI mappings, and DCBX profiles. Combined with IP Maestro for lifecycle management and gNMI streaming telemetry, OcNOS-DC enables fully automated Day-0 through Day-2 operations without manual CLI intervention.
What automation tools does OcNOS-DC support for DC fabric management?
OcNOS-DC is built for automation-first operations: NETCONF/YANG 1.1 and OpenConfig models for model-driven configuration, gNMI with dial-in and dial-out streaming telemetry for real-time visibility, Ansible playbooks for idempotent fabric-wide deployments, a Terraform provider for infrastructure-as-code workflows, and sFlow for traffic sampling. Docker and Kubernetes support in OcNOS 7.0 allows third-party monitoring agents to run directly on the switch CPU, eliminating the need for separate collector infrastructure.