NVIDIA MLNX_OFED Documentation v5.9-0.5.6.0.107
Linux Kernel Upstream Release Notes v5.17

Storage Protocols

There are several storage protocols that use the advantage of InfiniBand and RDMA for performance reasons (high throughput, low latency and low CPU utilization). In this chapter we will discuss the following protocols:

  • SCSI RDMA Protocol (SRP) is designed to take full advantage of the protocol off-load and RDMA features provided by the InfiniBand architecture.

  • iSCSI Extensions for RDMA (iSER) is an extension of the data transfer model of iSCSI, a storage networking standard for TCP/IP. It uses the iSCSI components while taking the advantage of the RDMA protocol suite. ISER is implemented on various stor- age targets such as TGT, LIO, SCST and out of scope of this manual.

    For various ISER targets configuration steps, troubleshooting and debugging, as well as other implementation of storage protocols over RDMA (such as Ceph over RDMA, nbdX and more) refer to Storage Solutions on the Community website.

  • Lustre is an open-source, parallel distributed file system, generally used for large-scale cluster computing that supports many requirements of leadership class HPC simulation environments.

  • NVM Express™ over Fabrics (NVME-oF)

    • NVME-oF is a technology specification for networking storage designed to enable NVMe message-based commands to transfer data between a host computer and a target solid-state storage device or system over a network such as Ethernet, Fibre Channel, and InfiniBand. Tunneling NVMe commands through an RDMA fabric provides a high throughput and a low latency. This is an alternative to the SCSi based storage networking protocols.

    • NVME-oF Target Offload is an implementation of the new NVME-oF standard Target (server) side in hardware. Starting from ConnectX-5 family cards, all regular IO requests can be processed by the HCA, with the HCA sending IO requests directly to a real NVMe PCI device, using peer-to-peer PCI communications. This means that excluding connection management and error flows, no CPU utilization will be observed during NVME-oF traffic.
      For further information, please refer to Storage Solutions on the Community website (enterprise-support.nvidia.com/s/).

© Copyright 2023, NVIDIA. Last updated on Nov 27, 2023.