Which technology is required on an Ethernet network to make RoCE traffic lossless?

Link Aggregation Control Protocol (LACP)

A network engineer is configuring a converged Ethernet fabric that will carry both storage (iSCSI) traffic and AI training (RoCEv2) traffic. The engineer needs to ensure that storage traffic is never dropped and that RoCE traffic gets a guaranteed share of bandwidth, while also preventing a single traffic class from consuming all available buffer space. Which combination of DCB features should be configured?

ETS only, with no PFC configured on any class

PFC on both iSCSI and RoCE classes, and ETS to allocate bandwidth between them

PFC on all classes and ECN disabled on all classes

PFC only on the RoCE class, and ECN marking on the iSCSI class

How do AI frameworks like Horovod or NCCL benefit from network features such as RoCE?

They can encrypt all traffic without performance impact.

They can replace the need for a network.

They can use RDMA to accelerate communication between GPUs.

They can automatically configure the network.

What is a key benefit of a spine-leaf (Clos) topology for AI data centers?

It reduces the number of switches needed compared to a traditional three-tier design.

It offers predictable, low-latency paths with high scalability.

It provides a single point of failure for simplicity.

It is only suitable for small networks.

Which statement best describes the network traffic pattern of a typical distributed AI training workload?

Traffic is predominantly north-south, with clients sending small requests to a centralized inference server.

Traffic is mostly southbound, with storage systems pushing data to compute nodes on a schedule.

Traffic is predominantly east-west, with large data transfers between compute nodes during gradient synchronization.

Traffic is evenly balanced between north-south and east-west, with each direction carrying equal volume.

CiscoCertified Network Professional Data Center

Domain 4Objective 3

4.3 Describe High-Performance Network Enabling Technologies for AI in Data Center Infrastructure 350-601 Practice Questions (Page 5)

Part of the Automation and Artificial Intelligence domain, which accounts for 15% of the 350-601 exam. Cisco does not publish an official question count, but from its 120-minute exam (~50–80 total, ~8–12 in this domain), expect 3–4 from this objective — we provide 31 practice questions to prepare you well beyond it. (estimate)

31questions here

7free pages

9concepts

15%of the exam

Questions 21–25

21foundation · easy
Which technology is required on an Ethernet network to make RoCE traffic lossless?
Select an answer first
22application · medium
A network engineer is configuring a converged Ethernet fabric that will carry both storage (iSCSI) traffic and AI training (RoCEv2) traffic. The engineer needs to ensure that storage traffic is never dropped and that RoCE traffic gets a guaranteed share of bandwidth, while also preventing a single traffic class from consuming all available buffer space. Which combination of DCB features should be configured?
Select an answer first
23foundation · easy
How do AI frameworks like Horovod or NCCL benefit from network features such as RoCE?
Select an answer first
24foundation · easy
What is a key benefit of a spine-leaf (Clos) topology for AI data centers?
Select an answer first
25foundation · easy
Which statement best describes the network traffic pattern of a typical distributed AI training workload?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to 350-601

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Cisco. “350-601” is a trademark of its owner, used for identification only.