Cisco Cisco Prime Virtual Network Analysis Module (vNAM) 6.2 White Paper

network. In this context, the network is the unified fabric. FCoE, VM-FEX, vPCs and FabricPath are 
Ethernet technologies that have evolved data center fabric design options. These technologies can be used 
concurrently over the VMDC Nexus-based infrastructure.

FCoE uses FSPF (Fabric Shortest Path First) forwarding, which FabricPath does not yet support 
(FabricPath uses an IS-IS control plane). FCoE must be transported on separate (classical Ethernet) 
VLANs. In VMDC VSA 1.0, we assume that FCoE links are leveraged outside of the FabricPath 
domain—such as within the ICS portions of the FabricPath-based pod—to reduce cabling and adapter 
expenses and to realize power and space savings.

The VMDC compute architecture assumes, as a baseline premise, a high degree of server virtualization, 
driven by data center consolidation, the dynamic resource allocation requirements fundamental to a 
"cloud" model, and the need to maximize operational efficiencies while reducing capital expense 
(CAPEX). Therefore, the architecture is based upon three key elements:

Hypervisor-based Virtualization—In VMDC VSA 1.0, as in previous VMDC releases, VMware 
vSphere plays a key role, logically abstracting the server environment in terms of CPU, memory, 
and network into multiple virtual software containers to enable VM creation on physical servers. In 
this release, vSphere VMs provide the foundation for router and service node virtualization.

Separate, interrelated documents address Microsoft Hyper-V and Nexus 1000V integration 
for application workloads in VMDC FabricPath systems: 

UCS Network, Server, and I/O Resources in a Converged System—UCS provides a highly 
resilient, low-latency unified fabric for integrating lossless 10 Gigabit Ethernet and FCoE functions 
using x86 server architectures. UCS provides a stateless compute environment that abstracts I/O 
resources, server personality, configuration, and connectivity to facilitate dynamic programmability. 
Hardware state abstraction simplifies moving applications and operating systems across server 
hardware.

The Nexus 1000V—This virtual switch, which provides a feature-rich alternative to VMware 
Distributed Virtual Switch, incorporates software-based VN-link technology to extend network 
visibility, QoS, and security policy to the VM level. VMDC VSA 1.0 uses VMware vSphere 5.1 as 
the compute virtualization operating system. A complete list of new vSphere 5.1 enhancements is 
available

 online.

 Key "baseline" vSphere features leveraged by the system include ESXi boot from 

SAN, VMware High Availability (HA), and Distributed Resource Scheduler (DRS). Basic to the 
virtualized compute architecture is the notion of clusters; a cluster comprises two or more hosts with 
their associated resource pools, VMs, and data stores. Working with vCenter as a compute domain 
manager, vSphere advanced functionality, such as HA and DRS, is built around the management of 
cluster resources. vSphere supports cluster sizes of up to 32 servers when HA or DRS features are 
used. In practice, however, the larger the scale of the compute environment and the higher the 
virtualization (VM, network interface, and port) requirements, the more advisable it is to use smaller 
cluster sizes to optimize performance and virtual interface port scale and limit the intra-cluster 
failure domain. Previously in VMDC large pod simulations, cluster sizes were limited to eight 
servers; in smaller pod simulations, cluster sizes of 16 or 32 were used. For VMDC VSA 1.0, cluster 
sizes of 16 servers are deployed in the system under test (SUT). As in previous VMDC releases,