NVIDIA-Certified Professional AI Networking

0/2 checked

Q_01 domain/d1 · medium single discuss →

An AI data center architect is designing an OVX system for distributed generative AI and industrial digitalization workloads and must select an east-west (E/W) fabric that can efficiently scale across hundreds of servers. Based on the OVX platform description, which combination of components enables ultra-fast, high-bandwidth communication for these distributed workloads?

A NVLink Switch fabric interconnecting the L40S GPUs within each server
B ConnectX-7 network adapters and BlueField-3 DPUs over low-latency Ethernet or InfiniBand
C MetroX long-haul systems extending connectivity between remote data centers
D PCIe Gen 6 host links between the CPUs and the L40S GPUs

Q_02 domain/d1 · medium single discuss →

A team is planning an InfiniBand fabric for a large AI/HPC cluster and wants to reduce the volume of data traversing the network during distributed training's collective communication phases. Which InfiniBand capability directly addresses this by offloading those operations to the switch network?

A Self-healing network capabilities
B Adaptive routing with congestion control (quality of service)
C Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)
D InfiniBand-to-Ethernet gateway systems

NVIDIA-Certified Professional AI Networking

Explanation

Explanation