The traditional Ethernet networks are optimized for best-effort traffic, tolerating frame loss, retransmission, packet collisions, and out-of-order delivery. While sufficient for most general-purpose data traffic, this behavior is inadequate for storage traffic, such as Fibre Channel over Ethernet (FCoE), which demands lossless transport across the network.

To address this challenge, the IEEE 802.1 Data Center Bridging (DCB) standard introduced enhancements that enable lossless, low-latency transmission of sensitive data types over Ethernet, making it suitable for converged data center networks.

Data Centre Bridging (DCB) is a set of enhancements for Ethernet that enables both LANs and Storage Area Networks (SANs) to utilize a single integrated infrastructure within a data center. The DCB technology enables the transportation of Fiber Channel, TCP/IP, and inter-process communication data across a unified Ethernet network.

DCB is also essential for AI/ML workload transport. For AI/ML workloads, much of the traffic, such as inference, control, gradient synchronization, activation, and feature map data traffic types are extremely sensitive to latency, bandwidth, and packet loss. Any compromise in these areas directly translates to increased Job Completion Time (JCT), reduced GPU utilization, and training instability. These traffic types are best transported via RoCEv2 over dedicated, L3 routed GPU-to-GPU fabric with ECMP. PFC over Layer 3 enables lossless Ethernet transport across Layer 3 spine-leaf topology-based CLOS fabrics.

Differentiated Traffic Handling

DCB allows multiple traffic types—such as storage, voice, and general data—to coexist and be managed differently on the same physical Ethernet link. This ensures that latency-sensitive or loss-intolerant traffic receives the appropriate level of service.

Lossless Ethernet with Flow Control

To minimize frame loss due to congestion, DCB includes mechanisms that control the flow of traffic at a granular level.

Limitations:

When a switchport (L2) is configured with Priority Flow Control (PFC), applying the no switchport command will also clear the PFC configuration. This behavior is part of the cleanup process triggered by the no switchport command.

DCBX is primarily used to manage the following DCB protocols:

Priority-based Flow Control (PFC) (IEEE 802.1Qbb) is a link-level mechanism that enables the selective pausing of traffic based on priority levels to prevent packet loss, offering granular control over how various traffic classes are managed. While traditionally deployed in Layer 2 (Ethernet) environments, extending PFC to Layer 3 interfaces brings these benefits to routed networks, enhancing traffic handling across complex topologies. By providing flow control per class, PFC helps minimize packet loss, particularly in congestion-sensitive applications such as storage, AI/ML, and high-performance computing.

PFC is a Layer 2 mechanism — it only works between directly connected neighbors on an Ethernet link. PFC alone can not operate end-to-end on a typical IP-routed (L3) network. However, when the packet is routed hop-by-hop at L3, the actual transmission on each link is L2. Therefore, PFC can still apply at each hop, controlling traffic for a given priority (e.g., for CoS 3 used by RoCEv2). The result is that lossless behavior is preserved link-by-link, provided all hops agree on which priorities to pause, even though the overall path is routed.

Large-scale model training such as deep learning models like GPT or ResNet involves high-volume east-west traffic between GPUs/TPUs and storage.

High throughput, low congestion, and lossless transport such as RoCEv2 with PFC.

Ensures lossless delivery for storage and real-time traffic.

Enhances network convergence by supporting multiple traffic types on one fabric.

Reduces infrastructure cost by eliminating the need for separate SAN and Ethernet networks.

The DCB supports the following configurations: