The illustrations in this section assume it is possible to connect (trunk) the same VLAN between the data center’s external connectivity environments and the workload compute environments. In some infrastructure topologies, this is not possible because multiple routing hops might ordinarily exist between different parts of the infrastructure. Solutions employing discrete rack-mount compute nodes (as opposed to blade-based solutions with integrated networking) are often configured in a “leaf-spine” topology where VLANs are constrained within a rack, and inter-rack connectivity is achieved over a routed Layer 3 core.

When there is no ubiquitous Layer 2 coverage across the entire compute estate, the placement of tenant Edge Services Gateways onto compute resources with access to the required external connectivity becomes a challenge. vCloud Director offers some assistance because its placement engine deploys edge VMs to clusters based upon VLAN connectivity. Presenting the VLANs directly to a subset of racks (typically at least two for resiliency) provides that Edge Services Gateways will only be deployed to those clusters because of the VLAN availability there. However, vCloud Director will also be able to provision ordinary workloads to those same clusters because of the availability of tenant networking required for the southbound connectivity from those same Edge Services Gateways. Because of this, clusters with external connectivity in this model are known as combined Edge/Compute clusters.

There are a number of alternative models for the placement of Edge Services Gateways on separate “Edge Clusters” without tenant workloads, each of which brings its own complexity and operational overhead. The principles outlined earlier in this section still apply, but must be applied on top of the Provider’s chosen Edge cluster option. The options are discussed in more detail in Architecting a VMware vCloud Director Solution for VMware Cloud Providers (Section 6.3.2). See the Section 7, References for more information.