Data Center Bridging

Data Center Bridging
Name	Data Center Bridging
Abbreviation	DCB
Developed by	IEEE, InfiniBand Trade Association, ISO
Initial release	2000s
Domain	Data center networking

Data Center Bridging Data Center Bridging is a collection of enhancements to Ethernet standards designed to provide lossless, low-latency, and traffic-managed fabric suitable for converged data center environments. It aims to enable unified transport for storage, virtualization, and high-performance computing by coordinating features that control congestion, prioritize flows, and allocate bandwidth across switches and adapters. The framework is closely associated with organizations such as the IEEE, Storage Networking Industry Association, and vendors like Cisco Systems, Intel Corporation, and Broadcom Inc..

Overview

Data Center Bridging defines an ecosystem of features that extend traditional Ethernet to meet requirements from communities such as Fibre Channel, InfiniBand, NVMe over Fabrics, Microsoft virtualization workloads, and large-scale deployments by Google, Facebook, Amazon Web Services, and Microsoft Azure. It addresses use cases driven by technologies like Fibre Channel over Ethernet, iSCSI, RDMA over Converged Ethernet, and converged storage and compute environments found in Hadoop clusters, OpenStack clouds, and Kubernetes orchestration. The initiative overlaps with standardization efforts from ISO, IETF, and the Broadband Forum where multi-vendor interoperability is critical.

Technical Components

Key components include Priority-based Flow Control, Enhancement to Transmission Selection, Data Center Bridging Exchange protocol, and Quantized Congestion Notification. Priority-based Flow Control (PFC) implements pause semantics per priority to protect loss-sensitive traffic such as Fibre Channel and RDMA; this mechanism is interoperable with NICs from Mellanox Technologies and ASICs by Broadcom Inc.. Enhanced Transmission Selection (ETS) provides bandwidth allocation among traffic classes and interacts with scheduler implementations on switches from vendors like Arista Networks and Juniper Networks. Quantized Congestion Notification (QCN) offers congestion signaling from switches to endpoints, and the Data Center Bridging Exchange (DCBX) protocol allows automatic configuration exchange between peers, a capability supported in firmware from Intel Corporation and management stacks from VMware, Inc..

Standards and Protocols

The suite maps to specific standards including IEEE 802.1Qbb for PFC, IEEE 802.1Qaz for ETS and DCBX, and IEEE 802.1Qau for QCN; these are coordinated with the IEEE 802 family and other bodies such as the IETF for congestion control. Related specifications include T11 initiatives for Fibre Channel and the InfiniBand Trade Association's RDMA over Converged Ethernet guidance. Interoperability profiles and conformance testing have been discussed at consortia like the Open Networking Foundation and the Storage Networking Industry Association, with vendor compliance often referenced in interoperability events hosted by InterOp-style gatherings.

Implementation and Deployment

Deployments require support across switches, adapters, and host stacks; major data center operators such as Google and Facebook evaluate implementations from Cisco Systems, Arista Networks, Juniper Networks, and hyperscaler in-house designs. Host drivers implementing RDMA over Converged Ethernet are maintained in the Linux kernel and by Microsoft for Windows Server, and vendor firmware from Mellanox Technologies and Intel Corporation implements DCBX TLV exchange. Integration with orchestration platforms like OpenStack, Kubernetes, and VMware vSphere involves coordination of QoS policies, and validated reference architectures have been published by Dell Technologies and Hewlett Packard Enterprise.

Performance and Compatibility

When correctly configured, the features reduce packet loss for storage and RDMA workloads and can lower application latency in environments running SAP HANA, large-scale MySQL farms, and HPC clusters using MPI. However, mixed-vendor environments often require careful tuning because implementations of PFC, ETS, and QCN differ among ASICs from Broadcom Inc., Marvell Technology, and Intel Corporation. Performance trade-offs are studied in papers from academic institutions such as Massachusetts Institute of Technology, Stanford University, and University of California, Berkeley, and in benchmarking by vendors and industry groups like the Storage Networking Industry Association.

History and Development

The effort emerged in response to demands from storage and virtualization communities in the early 2000s and matured through IEEE ballot cycles in the 2000s and 2010s. Stakeholders included equipment manufacturers such as Cisco Systems and Brocade Communications Systems and hyperscalers including Google and Amazon Web Services. Research contributions and early prototypes were produced at laboratories and universities like IBM Research, Intel Labs, Microsoft Research, and academic groups at Carnegie Mellon University and University of Illinois Urbana–Champaign.

Security and Management

Operational security and management practices involve firmware validation by vendors including Intel Corporation and Broadcom Inc., network management via systems from SolarWinds, Cisco Prime, and orchestration via Ansible and Puppet. Misconfiguration of PFC or ETS can create denial-of-service conditions affecting storage fabrics used by Oracle databases or virtual workloads in VMware vSphere, so monitoring and telemetry—leveraging protocols such as SNMP and streaming telemetry initiatives from the IETF—are essential. Industry working groups such as the Storage Networking Industry Association and the Open Compute Project publish best practices for secure deployment and management.

Category:Computer networking