Data Center Networks Design Hub

Network design resources, include Nokia Validated Designs (NVD) and Nokia Reference Design documents.

Nokia Validated Designs (NVDs) address the challenges of complex multivendor data center integration, long deployment cycles, and costly downtime and human error by removing uncertainty and simplifying infrastructure deployment.
Nokia Reference Designs are guides that showcase alternate design and product capabilities of the Nokia portfolio. These designs are deployed as virtual fabrics using containerlab. They have deployable GitHub repositories that can be downloaded and customized for further use, testing, and exploration.

Document name	Document type	Description	Resources
3-stage EVPN/VXLAN	Validated Design	This validated design document is based on a 3-stage Clos EVPN/VXLAN architecture, detailing physical and logical connectivity aspects, as well as associated technologies, for a single-site, multi-tiered data center deployment. The design leverages EVPN as the control plane and VXLAN as the data plane.	Validated Design GitHub repository
Collapsed spine EVPN /VXLAN	Validated Design	This document outlines a collapsed spine EVPN/VXLAN design, describing various physical and logical connectivity aspects and associated technologies involved in a single- site, single-tiered data center architecture, with EVPN as the control plane and VXLAN as the data plane.	Validated Design GitHub repository
4-way collapsed spine EVPN/VXLAN	Reference Design	This collapsed spine reference design demonstrates an architecture with enhanced scale-out requirements, using four spines instead of the typical two, while still following the general principles of a collapsed spine design.	Reference Design GitHub repository
Congestion management in AI/ML networks	Technical brief	This document explains congestion detection, management, and avoidance in AI/ML networks. When combined with collision-free design principles, congestion management enables a lossless AI fabric essential for AI/ML.	Technical Brief
Lenovo-Nokia AI-DC Validated Design	Validated design	This document provides a validated design for deploying AI/ML data center solutions using a best-in-class combination of networking, compute, storage, and GPU technologies. Developed through close collaboration between leading technology vendors, the solution has been designed and tested to meet the demanding requirements of modern AI workloads across training, inference, and data processing use cases.	Validated design GitHub repository

Document name

Document type

Description

Resources

3-stage EVPN/VXLAN

Validated Design

This validated design document is based on a 3-stage Clos EVPN/VXLAN architecture, detailing physical and logical connectivity aspects, as well as associated technologies, for a single-site, multi-tiered data center deployment. The design leverages EVPN as the control plane and VXLAN as the data plane.

Validated Design

GitHub repository

Collapsed spine EVPN /VXLAN

Validated Design

This document outlines a collapsed spine EVPN/VXLAN design, describing various physical and logical connectivity aspects and associated technologies involved in a single- site, single-tiered data center architecture, with EVPN as the control plane and VXLAN as the data plane.

Validated Design

GitHub repository

4-way collapsed spine EVPN/VXLAN

Reference Design

This collapsed spine reference design demonstrates an architecture with enhanced scale-out requirements, using four spines instead of the typical two, while still following the general principles of a collapsed spine design.

Reference Design

GitHub repository

Congestion management in AI/ML networks

Technical brief

This document explains congestion detection, management, and avoidance in AI/ML networks. When combined with collision-free design principles, congestion management enables a lossless AI fabric essential for AI/ML.

Technical Brief

Lenovo-Nokia AI-DC Validated Design

Validated design

This document provides a validated design for deploying AI/ML data center solutions using a best-in-class combination of networking, compute, storage, and GPU technologies. Developed through close collaboration between leading technology vendors, the solution has been designed and tested to meet the demanding requirements of modern AI workloads across training, inference, and data processing use cases.

Validated design

GitHub repository

| Doc Center