KB 915003 - Enabling and Validating RoCE (RDMA over Converged Ethernet) on SONiC

KB 915003 - Enabling and Validating RoCE (RDMA over Converged Ethernet) on SONiC

 Purpose

This guide explains how to enable RoCE on a SONiC switch, verify RDMA interfaces on connected compute nodes, and validate performance using RDMA bandwidth tests.

 1. Enable RoCE on SONiC Switch

Step 1.1 — Enable Priority Flow Control (PFC)

PFC ensures lossless Ethernet behavior required by RoCE.

sonic-cli
configure terminal
# Enable RoCe with default settings
sonic(config)# roce enable
  force-defaults  Clear any previous applied QOS buffer and force Initialize RoCEv2 default buffer configuration
  pfc-priority    RoCEv2 buffer configuration based on the PFC priorities
  <cr>

sonic(config)# roce enable
This command will also restart the node after saving all configurations,if ROCE is configured first time or force-default. [Proceed y/N]: y
sonic(config)# Waiting for the reboot operation to complete






Step 1.4 — Verify RoCE-Ready QoS Configuration

Check that PFC is active and watchdog is off (optional):

show qos interface Ethernet0

Expected output:

sonic# show qos interface Ethernet 0
          scheduler policy: ROCE
          dscp-tc-map: ROCE
          dot1p-tc-map: ROCE
          tc-queue-map: ROCE
          tc-pg-map: ROCE
          pfc-priority-queue-map: ROCE
          pfc-priority-pg-map: ROCE
          pfc-asymmetric: off
          pfc-priority  : 3,4
          PFC Watchdog
            Status            : on
            Action            : drop
            Detection Time    : 200ms
            Restoration Time  : 400ms

Result: SONiC switch ports are now RoCE-ready.

 2. Verify RDMA on Compute Nodes

Run these on each Ubuntu compute node.

Step 2.1 — Check RDMA Interfaces

rdma link show

Expected output:

link rocep65s0f0/1 state ACTIVE physical_state LINK_UP netdev enp65s0f0np0
link rocep65s0f1/1 state ACTIVE physical_state LINK_UP netdev enp65s0f1np1
ibstat

Expected output:

Link layer: Ethernet
Rate: 200
State: Active

Result: Indicates that RDMA (RoCE) links are active and running on Ethernet mode at 200 Gb/s.

 3. Perform RoCE Bandwidth Test Between Nodes

Test Setup

VLAN Purpose Node 1 IP Node 2 IP Interface
VLAN 100 HSN-A 10.1.1.11 10.1.1.12 rocep65s0f1 / rocep1s0f0
VLAN 200 HSN-B 10.1.2.11 10.1.2.12 rocep65s0f0 / rocep1s0f1

Make sure RDMA utilities are installed:

sudo apt install rdma-core ibverbs-utils perftest -y

Step 3.1 — Run Bandwidth Test (HSN-A / VLAN 100)

On Node 1 (Server)

ib_write_bw -d rocep65s0f1 -F --report_gbits

On Node 2 (Client)

ib_write_bw -d rocep1s0f0 -F --report_gbits 10.1.1.11

Expected output:

---------------------------------------------------------------------------------------
 #bytes     #iterations    BW peak[Gb/sec]    BW average[Gb/sec]   MsgRate[Mpps]
 65536      5000             185.05             185.03             0.352922
---------------------------------------------------------------------------------------

Result: Confirms 185+ Gb/s effective throughput on 200G RoCE link.


 4. Validate Performance Counters


Check RDMA statistics:

rdma statistic show


    • Recent Articles

    • KB-630208 | How to use Find and Locate to search for files in Linux

      Purpose To provide Linux administrators and Field Engineers with guidance on using the find and locate utilities to efficiently search for files and directories within a Linux filesystem based on criteria such as name, type, size, timestamps, ...
    • KB-630164 | How to use rsync to Synchronize Files

      Purpose To provide a clear and practical procedure for using the rsync utility to synchronize files and directories between local systems and remote hosts, including push and pull operations, while preserving file attributes and optimizing data ...
    • KB-630107 | How to create a Linux swap file

      Purpose To provide Field Engineers and Linux administrators with a standardized procedure for creating, configuring, and validating a Linux swap file. This procedure helps ensure sufficient virtual memory is available when physical RAM is exhausted ...
    • MSI All-in-One (AIO) PC – Physical Damage Policy for Warranty Claims

      MSI All-in-One (AIO) PC – Physical Damage Policy for Warranty Claims MSI's standard warranty does not cover physical or accidental damage to an All-in-One (AIO) PC. If the damage is determined to be caused by external factors rather than a ...
    • KB 134833 - Troubleshooting NVSM Alert NV-CPU-XX – Unrecoverable CPU Internal Error

      Purpose This document provides a general troubleshooting procedure for the NVIDIA System Management (NVSM) alert NV-CPU-XX, which indicates that a CPU has reported an internal error. The article outlines how to verify whether the alert represents an ...
    • Popular Articles

    • CP Plus Camera and NVR Configuration

      NVR Configuration The CP Plus Pro Series of NVRs have been meticulously designed for providing you with upgraded performance and higher recording quality in your IP video surveillance solution. The robust processor that has been inculcated in this ...
    • KB 775235 - Kerberos Authentication – Overview

      What is Kerberos? Kerberos is a secure authentication method used in our Active Directory (AD) environment (mbuzztech.com). It allows users to: Access multiple systems without re-entering passwords (Single Sign-On – SSO) Log in once Where We Use It ...
    • How to Remove and Reinstall NVIDIA Drivers on Ubuntu

      This article provides step by step guide to completely remove existing NVIDIA drivers and reinstall specific version of the NVIDIA driver on the Ubuntu system Prerequisites Administrative (sudo) access to the Ubuntu system. Internet access to ...
    • KB 692001 - Personal Computers and Servers - Classification and Point of Contact

      We can classify the computers that MBUZZ handles based on their form-factor as below: Tower Workstations, Desktops, Gaming PCs and SFF (Small form factor) PCs fall under this category. These are computers people would use on a desk and rarely move. ...
    • KB 298031 - M.2 SSD Tier List

      The sequential read and write speeds, which are usually the most advertised number, are not a proper benchmark of real-world performance or the quality of an SSD. This article categorizes and tiers SSDs based on factors like the type of NAND flash, ...