To identify the activeSubnet Manager (SM)node in an InfiniBand fabric, the correct command sequence is:

sminfo

Displays general information about the active SM in the fabric, including itsLID.

smpquery ND <LID>

Resolves theNode Description (ND)at the given LID, revealing the exact hostname or label of the SM server.

From the InfiniBand Tools Guide:

"The sminfo utility provides the LID of the master SM. Use smpquery ND <LID> to resolve the node name hosting the SM."

This two-step approach is standard for locating and validating the SM identity in fabric diagnostics.

Incorrect Options:

B (Nl)is an invalid query type.

CandDdo not identify SMs.

Modern AI training (especially with LLMs) requires extremely high-speed, parallel access to large datasets. Adedicated storage fabricseparates data I/O traffic from the training compute path and avoids contention.

FromNVIDIA DGX Infrastructure Reference Architectures:

"Dedicated storage networks eliminate I/O bottlenecks by providing low-latency, high-bandwidth access to distributed storage for large-scale training jobs."

"Parallel access to datasets is key for performance, especially in multi-node, multi-GPU AI clusters."

Security (B)is important, but not the core reason for a storage fabric.

Cost (D)is typicallyincreased, not reduced, with dedicated fabrics.

To upgrade the MLNX-OS on InfiniBand switches, the recommended procedure is as follows:​

Connect to the switch via SSH: Establish a secure shell connection to the switch using its management IP address.​

Fetch the MLNX-OS software image: Obtain the appropriate MLNX-OS software image from the official source or repository.​

Use the 'install' command to perform the upgrade: Execute the 'install' command on the switch to initiate the upgrade process with the fetched software image.​

This method ensures a smooth and efficient upgrade without the need for physical intervention or service disruption.​

Reference Extracts from NVIDIA Documentation:

"Click on Systems → MLNX-OS Upgrade. Select the desired upgrade method (e.g. 'Install from local file'). Select your image and click 'Install Image'."

In Cumulus Linux, configuring a bond interface with Link Aggregation Control Protocol (LACP) involves setting the bond mode to 'lacp'. The correct command to achieve this is:​

nv set interface bond1 bond mode lacp​

This command sets the bonding mode of 'bond1' to LACP, enabling dynamic link aggregation for increased bandwidth and redundancy.​

Reference Extracts from NVIDIA Documentation:

"To reset the link aggregation mode for bond1 to the default value of 802.3ad, run the nv set interface bond1 bond mode lacp command."​

To check the status and link layer of InfiniBand interfaces, the ibstat command is used. For example:​

ibstat -d mlx5_0

This command provides detailed information about the InfiniBand device, including its state (e.g., Active), physical state (e.g., LinkUp), and link layer (e.g., InfiniBand).​

NVIDIA NetQ is a highly scalable network operations toolset that provides visibility, troubleshooting, and validation of networks in real-time. It delivers actionable insights and operational intelligence about the health of data center networks—from the container or host all the way to the switch and port—enabling a NetDevOps approach.

NetQ can be used as the functional test platform for the network CI/CD in conjunction with NVIDIA Air. Customers benefit from testing the new configuration with NetQ in the NVIDIA Air environment (“digital twin”) and fix errors before deploying to their production. ​

To enable RDMA over Converged Ethernet (RoCE) on a Cumulus switch, the correct command is:​

nv set qos roce enable on

This command configures the Quality of Service (QoS) settings to support RoCE, ensuring that the necessary parameters for lossless Ethernet are applied.​