Skip to content
Ask AI

Data Center & AI Fabric

Data Center & AI Fabric documents AsterNOS on CX-N and related data-center switches: leaf–spine underlay, BGP EVPN VxLAN overlay, MC-LAG, ECMP, and lossless Ethernet for AI training (RoCEv2, PFC, ECN, and Easy RoCE).

Operators use a Cisco-like CLI (KLISH) to configure fabrics that scale from small GPU clusters to multi-pod AI backends. The library covers installation, day-one access, feature configuration, REST API automation, and field-proven technical designs.

CapabilityTypical use
BGP EVPN VxLANMulti-tenant L2/L3 overlay, anycast gateway, symmetric IRB
MC-LAG + unnumbered BGPDual-homed servers and resilient server access
RoCEv2 / PFC / ECNLossless AI backend and storage networks
Telemetry & SNMPStreaming counters, buffers, and fabric health
REST APIController and OSS integration for fabric automation
SectionStart here
InstallationHardware compatibilityONIE preparationAsterNOS installation
First loginFirst use — console, SSH, and basic management
ConfigurationConfiguration guide preface — feature chapters in the left sidebar
CLI referenceCommand-line reference — per-feature syntax
REST APIREST API manual — programmatic fabric operations
Design guidesTechnical docs — AI fabric and EVPN MC-LAG best practices
  1. Confirm platform support in the hardware compatibility list.
  2. Install AsterNOS with the installation guide.
  3. Build underlay routing and VxLAN overlay — see VxLAN configuration and BGP.
  4. For AI backends, follow intelligent lossless network and the small-scale or medium/large-scale fabric guides.