Understanding FCoE (Fibre Channel over Ethernet) and Its Standards

FCoE originated from a proposal by Brocade, IBM, and HP called FCoCEE, later renamed Fibre Channel over Ethernet, with 16 major vendors (Cisco, EMC, Emulex, HDS, HP, IBM, Intel, LSI, Marvell, Mellanox, Neoscale, Nuova, PMC‑Sierra, Qlogic, Sun, Vitesse) participating in its development.

The key standards are T11.3‑FC‑BB‑5 (finalized June 2009) and T11.3‑FC‑BB‑6 (finalized August 2009), the latter defining a point‑to‑point, end‑to‑end FCoE model.

FC‑BB‑5 requires an FCF (FC Forwarder), essentially an FCoE switch, to replace the direct HBA‑to‑storage connections used in traditional FC. An optional FSB device can extend ports and supports FIP (FCoE Initialization Protocol) snooping.

After FC‑BB‑6, point‑to‑point networking is possible, and VN2VN topologies can be built using inexpensive FDF devices.

FCoE (Fibre Channel over Ethernet) maps the FC architecture onto Ethernet, leveraging Ethernet’s maturity and speed while preserving existing FC investments and reducing cabling and interface cards in data‑center environments.

FCoE encapsulates FC frames within Ethernet frames, using extensions to the 802.1Q standard (often referred to as 802.1Qxx or 802.3) to ensure lossless transmission.

Enhanced Ethernet (CEE/DCE) is required for lossless transport; Data Center Bridging (DCB) technologies such as Priority‑based Flow Control (PFC), Enhanced Transmission Selection (ETS), Congestion Notification (CN), and DCBX provide the necessary flow‑control mechanisms.

Traditional Ethernet drops packets under congestion, which FC cannot tolerate; therefore, DCB modifies Ethernet to use lossless mechanisms (PFC, ETS, CN) to guarantee uninterrupted FC traffic.

The main benefit is simplified cabling on the server side: servers use a Converged Network Adapter (CNA) to connect to an FCoE switch, halving the number of cables, while storage can remain FC‑connected to the same switch.

Key enhanced Ethernet features supporting FCoE include full‑duplex operation, ordered delivery, and jumbo frames (minimum 2.5 KB payload per frame).

PFC (IEEE 802.1Qbb) enables priority‑based pause frames, preventing one traffic class from starving others; ETS (IEEE 802.1Qaz) allocates bandwidth among virtual links created for PFC; DCBX exchanges Ethernet parameters between switches and endpoints; CN (IEEE 802.1Qau) notifies congestion to limit bandwidth at the edge.

The FCoE protocol model retains FC‑4 (high‑level protocol mapping), FC‑3 (common services), and FC‑2 (FC frames), while FC‑0/1 are replaced by Ethernet’s physical and MAC layers, with a mapping layer translating between FC and Ethernet.

Port naming differences: N_Port (HBA) becomes VN_Port (CNA), F_Port (fabric) becomes VF_Port (FCoE switch port), and E_Port (FC‑to‑FC) becomes VE_Port (inter‑FCoE‑switch link).

FCoE networking types include VN2VF (ENODE to FCF), VE2VE (FCF to FCF), and VN2VN (ENODE to ENODE) topologies, each illustrated with diagrams showing physical lossless Ethernet connections and virtual FCoE links.