Deployment Guide

10 ScaleIO/VxFlex OS IP Fabric Best Practice and Deployment Guide with OS10EE | version 1.0
2 Building a Leaf-Spine topology
This paper describes a general-purpose virtualization infrastructure suitable for a modern data center. The
solution is based on a Leaf-Spine topology utilizing Dell EMC Networking S4248FB-ON switches for the leaf
switches, and Dell EMC Networking Z9100-ON switches for the spine switches. This example topology uses
four spine switches to maximize throughput between leaf switches and spine switches.
Figure 3 shows a high-level diagram of the Leaf-Spine topology used in this guide. As a best practice, each
new rack added to the data center contains two leaf switches. Join these two switches using a VLT
interconnect (VLTi) so the other layer 2 downstream devices see them as a single logical device.
The connections between spine switches and leaf switches can be layer 2 (switched) or layer 3 (routed). The
deployment scenario in this guide uses layer 3 connections. The main benefit is layer 2 broadcast domains
are limited, resulting in improved network stability and scalability.
The Dell EMC Networking Z9100-ON supports a maximum of 32 leaf switches per pod. However, the
example in this document uses only two leaf switches. A single rack can provide VxFlex OS management,
VMware management, and several VxFlex OS compute/storage nodes. As administrators add racks to the
data center, they add two leaf switches to each rack. As bandwidth requirements increase, administrators add
spine switches.
Leaf-Spine topology example