Administrator Guide

VLT on Core Switches
Uplinks from servers to the access layer and from access layer to the aggregation layer are bundled in LAG groups with end-to-end Layer
2 multipathing. This set up requires “horizontal” stacking at the access layer and VLT at the aggregation layer such that all the uplinks
from servers to access and access to aggregation are in Active-Active Load Sharing mode. This example provides the highest form of
resiliency, scaling, and load balancing in data center switching networks.
The following example shows stacking at the access, VLT in aggregation, and Layer 3 at the core.
Figure 131. VLT on Core Switches
The aggregation layer is mostly in the L2/L3 switching/routing layer. For better resiliency in the aggregation, Dell EMC Networking
recommends running the internal gateway protocol (IGP) on the VLTi VLAN to synchronize the L3 routing table across the two nodes on
a VLT system.
Enhanced VLT
Enhanced VLT (eVLT)) refers to the ability to connect two VLT domains. An eVLT configuration creates a port channel between two VLT
domains by allowing two different VLT domains, using different VLT domain ID numbers, connected by a standard link aggregation control
protocol (LACP) LAG to form a loop-free Layer 2 topology in the aggregation layer.
This configuration supports a maximum of four switches, increasing the number of available ports and allowing for dual redundancy of the
VLT. The following example shows how the core/aggregation port density in the Layer 2 topology is increased using eVLT. For inter-VLAN
routing, you do not need a separate router.
If you enable peer routing in an eVLT topology, a VLT node acts as a proxy gateway for its peer within the VLT domain. You can also
configure the two VLT domains to act as proxy gateways for each other. For more details, see the VLT Proxy Gateway chapter.
844
Virtual Link Trunking (VLT)