Traffic within the system is transported between Transport Nodes and devices attached to the Transport Nodes (such as Virtual Machines) by means of an encapsulation overlay. The overlay is a GENEVE tunnel built on the VxBlock physical network (Underlay). (GENEVE, Generic Network Virtualization Encapsulation, is a flexible network virtualization standard and encapsulation protocol.) Accessibility between devices attached to the overlay is managed by the Central Control Plane (CCP) and the Local Control Plane (LCP).

At some point:

1. Traffic will need to leave the system to reach devices/services external to NSX-T on the VxBlock 1000.
2. Devices external to the VxBlock 1000, and not participating in the overlay, will require ingress into the system.
3. Devices and endpoints must attach to centralized services that are contained within a central location within the system, such as stateful in-line firewalling.

The NSX-T Edge cluster performs this process of tunnel encapsulation/de-encapsulation, north-south ingress/egress, and provisioning of centralized services.

NSX-T on VxBlock 1000 offers a scalable and flexible yet prescriptive implementation of the NSX-T Edge cluster. An initial small footprint of two physical servers offers performance without being cost-prohibitive. The initial offering scales up to 16 servers, for customers who wish to implement additional centralized services at the edge.

Figure 5.         Scaling Edge Cluster

Edge Nodes

Edge Nodes are an NSX-T construct that are deployed on the physical servers. These Edge Nodes can be deployed in a virtual (VM-based) or physical (baremetal-based) form factor. For the initial VxBlock 1000 implementation of NSX-T, Dell EMC supports only virtual Edge Nodes. This provides an initial benefit in terms of deployment flexibility and serviceability. Future releases of NSX-T on VxBlock 1000 may integrate a bare-metal form factor to facilitate high performance use cases.