Fundamentals 12 min read

Understanding NVMe over Fabrics: Architecture, Transport Options, and Key Features

The article provides a comprehensive overview of NVMe over Fabrics, detailing its RDMA and TCP transport options, NVM subsystem architecture, namespace sharing, multi‑path I/O, reservation mechanisms, and industry adoption, highlighting its performance benefits for AI, ML, and real‑time analytics workloads.

Architects' Tech Alliance
Architects' Tech Alliance
Architects' Tech Alliance
Understanding NVMe over Fabrics: Architecture, Transport Options, and Key Features

NVMe over Fabrics (NVMe‑oF) extends the NVMe protocol beyond PCIe by using RDMA, Fibre Channel (FC), or native TCP transports, allowing high‑performance storage access over Ethernet, InfiniBand, and other fabrics.

The NVM subsystem consists of one or more controllers, ports, and namespaces; each controller typically connects to a single port, and ports may support only one transport type.

Namespace sharing enables multiple hosts to access a common namespace through separate controllers, while multi‑path I/O provides two or more independent paths between a host and a namespace for redundancy and load balancing.

NVMe‑oF includes discovery services that let hosts enumerate accessible NVM subsystems and their namespaces, and it supports reservation mechanisms similar to SCSI persistent reservations for coordinated access.

Recent developments introduce NVMe‑oF/TCP, offering a transport‑agnostic option that leverages existing Ethernet infrastructure without requiring RDMA, and the industry has embraced NVMe‑oF with support from major vendors such as Dell EMC, Intel, Broadcom, Cisco, and IBM.

NVMe‑oF delivers lower latency and higher throughput for AI, machine learning, and real‑time analytics workloads, and it simplifies storage stack integration by shortening the path between servers and NVMe SSDs.

storageNetworkingNVMedata centerNVMe-oFFabric
Architects' Tech Alliance
Written by

Architects' Tech Alliance

Sharing project experiences, insights into cutting-edge architectures, focusing on cloud computing, microservices, big data, hyper-convergence, storage, data protection, artificial intelligence, industry practices and solutions.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.