Cisco Cisco Prime Virtual Network Analysis Module (vNAM) 6.3 Libro bianco

three compute profiles are created to represent large, medium, and small application workloads: 
“Large” has 1 vCPU/core and 16 GB RAM; “Medium” has 0.5 vCPU/core and 8 GB RAM; and 
“Small” has 0.25 vCPU/core and 4 GB of RAM.

The Nexus 1000V provides additional logical segmentation capabilities using VXLANs. A 
MAC-in-UDP encapsulation, VXLAN packets feature a 24-bit LAN segment identifier that 
significantly increases logical scale in the infrastructure. The Nexus 1000V performs VXLAN 
encapsulation and de-encapsulation, so VXLANs are transparent to infrastructure layers north of 
this virtual access edge device.

Finally, the compute layer of the infrastructure can include bare metal servers for applications that are 
unsuitable for virtualization. In VMDC VSA 1.0, bare metal servers are attached via 1 GE interfaces to 
FEX 2200s attached to Nexus 5500 or Nexus 7000 access-edge (leaf) nodes.

The UCS-based compute architecture has the following characteristics:

It comprises multiple UCS 5100 Series chassis, each populated with eight half-width server blades.

Each server has dual 10 GigE attachments – in other words, to redundant A and B sides of the 
internal UCS fabric.

UCS is a fully redundant system, with two 2200 Series FEX per chassis and two 6200 Series Fabric 
Interconnects per system.

Internally, eight uplinks per FEX feed into dual Fabric Interconnects to pre-stage the system for the 
maximum possible bandwidth per server. This configuration means that each server has 20 GigE 
bandwidth for server-to-server traffic in the UCS fabric.