Cisco Unified Fabric
Historically, data centers have always had two separate major communication infrastructures: the local area network (LAN) and the storage area network (SAN).
The LAN infrastructure provides connectivity between the users and the applications running on the servers, and between the various applications in the data center, such as databases, application front ends, and so on. LAN infrastructures are used for the communication between computers using different network protocols. The most popular ones nowadays are the protocols in the Open System Interconnection (OSI) layer. The components of a LAN are the switches and routers, which build the infrastructure itself, and the computers connected to them. The computers need to have a network interface card (NIC), which is the LAN I/O interface used to connect and communicate on the network. The NICs support the communication of the OSI model protocols, and they operate at different speeds, starting with 10/100/1000Mbps and reaching 10/40Gbps and even 100Gbps with the latest Cisco adapters.
The LAN infrastructure is usually consolidated, which in this case means that, usually, unless there are security or compliance requirements, the LAN is one infrastructure, although it might be segmented at the different layers of the OSI model through the use of VLANs, routing, and security access-control lists. It can grow and it has different physical layers, but it stays one infrastructure.
There are also SAN infrastructures in the data center. SANs, as already covered in the previous chapters, are used for the communication between the servers and the storage systems. The servers and the storage systems need to be equipped with I/O controllers that support the communication of the Fibre Channel Protocol. These controllers are called host bus adapters (HBAs) and can communicate at speeds of 1/4/8/16 and 32Gbps. The SAN infrastructures are built by using Cisco MDS switches, which are specialized switches supporting the communication of the Fibre Channel Protocol. Historically, before Cisco entered the SAN market, the only security mechanism supported for isolating the communication was zoning. However, this mechanism was not enough, as it didn’t support true segmentation, and this led to the need for building multiple separate physical SAN infrastructures to separate and secure the data communication between different applications and their storage.
Cisco’s implementation of the Fibre Channel Protocol introduced the concept of virtual SANs (VSANs), which allowed for consolidating multiple separate physical SAN infrastructures, and on the top of a single SAN, separate virtual SANs, totally isolated from each other, could be created. This led to significant savings when organizations had to plan for and deploy storage infrastructures in their data centers. However, there was still the issue of having the two separate major infrastructures, the LAN and the SAN, each built with different types of devices, requiring the servers to have both NICs and HBAs. When specialized cabling is added into the budget, the cost becomes significant, especially with the price of the small form-factor pluggable (SFP) transceivers used for Fibre Channel Protocol connectivity. The SAN in general is much more expensive to be built compared to the cost of a LAN infrastructure. Figure 15-1 illustrates separate LAN and SAN infrastructures in the data center.

Figure 15-1 Separate LAN and SAN Infrastructures
To overcome these challenges and to offer a more flexible way of designing and deploying the needed communication infrastructures in the data center, Cisco came up with a new approach by introducing the creation of a consolidated I/O infrastructure, better known as the Cisco Unified Fabric. The idea is that the cheaper LAN infrastructure can be modified, using the Data Center Bridging enhancements, to create a lossless environment that allows the use of overlay encapsulation for the secure communication of the Fibre Channel Protocol frames. The Fibre Channel frames are encapsulated in Ethernet frames and sent over the Cisco Unified Fabric switches to the storage systems. The Cisco switches that support the FCoE protocol, and thus the Cisco Unified Fabric, include some of the Cisco Nexus and MDS switches as well as some of the Cisco Nexus 2000 fabric extenders. At the time this chapter was written, the list of Cisco switch models that support FCoE was constantly changing, so my advice would be to go and check with the Cisco product documentation for the latest information. Figure 15-2 illustrates the Cisco Unified Fabric.

Figure 15-2 Cisco Unified Fabric
In the Cisco Unified Fabric, the servers must use a different I/O controller that supports both the Ethernet and the FCoE protocols. These adapters are called converged adapters. The Cisco converged adapters offer extended capabilities compared to the ones offered by Broadcom and Emulex and are called Cisco virtual interface cards (VICs). The FCoE adapters are discussed in a dedicated section later in this chapter.
With the FCoE adapter, the server connects to the FCoE-capable switches with Ethernet cabling and also allows for FCoE communication over the same cheaper cabling. Although the FCoE standard does not impose any specific minimum bandwidth for the link, Cisco requires at least 10Gbps of bandwidth in order to support FCoE in a reliable manner.
The Cisco Unified Fabric provides the following benefits and flexibility in the data center:
- Reduced number of SAN switches.
- Reduced number of I/O adapters in the servers. Instead of separate NICs and HBAs, now converged adapters are used.
- Reduced cost of the cabling, as there are fewer or no native Fibre Channel links, which means less-expensive Fibre Channel Protocol transceivers.
- Significant savings in the amount of power per rack when FCoE is employed.
- Significant savings on power because less cooling is needed to compensate for the heat produced.
- Better cable management.
- Optimized device administration, as fewer devices are managed.
- Optimized troubleshooting, as fewer devices are involved in the communication path.
- Storage administrators manage their fabrics in the same manner they always have, as the FCoE does not change the Fibre Channel Protocol model.
- FCoE maps the Fibre Channel traffic onto lossless Ethernet. This results in performance benefits over technologies that require a gateway.
- FCoE is an extension of Fibre Channel onto the lossless Ethernet fabric.