NVIDIA Certified Professional: AI Networking Practice Questions: Spectrum-X Configuration Domain

Correct Answer: D

Explanation: Enabling Data Center Bridging (DCB) is a recommended practice in AI networks to ensure minimal latency and maximum throughput. DCB provides a set of enhancements for Ethernet networks, including Priority Flow Control (PFC) and Enhanced Transmission Selection (ETS), which help maintain a lossless network environment crucial for AI workloads. While QoS and ECMP can optimize traffic and load balancing, they do not directly address the need for a lossless environment. Port Mirroring is a diagnostic tool and does not contribute to performance optimization.

Correct Answer: C

Explanation: STP (Spanning Tree Protocol) is used to prevent network loops by creating a loop-free topology in Ethernet networks. This is essential for ensuring redundancy and reliability in enterprise AI networks. RIP, OSPF, and BGP are routing protocols and do not serve the purpose of preventing loops.

Correct Answer: B

Explanation: Quality of Service (QoS) is the feature that allows for the prioritization of AI traffic over other types of network traffic. By configuring QoS, network administrators can ensure that critical AI workloads receive the necessary bandwidth and low latency. NAT is used for IP address translation, port mirroring is for monitoring, and ACLs are for traffic filtering, not prioritization.

Correct Answer: B

Explanation: Implementing RADIUS authentication enhances security by providing centralized authentication, which is a common requirement for compliance with industry standards. SNMPv1 (A) is outdated and insecure. Allowing telnet access (C) is insecure due to lack of encryption. Disabling logging (D) reduces the ability to audit and ensure compliance.

Correct Answer: A

Explanation: The 'show vlan brief' command provides a concise summary of all VLANs configured on the switch, including their status and associated ports. This is essential for validating network segmentation in an enterprise AI environment. Options B, C, and D are not valid commands for checking VLAN configurations on Spectrum-X switches.

Correct Answer: B

Explanation: Using a centralized management tool like NVIDIA NetQ allows for efficient configuration, monitoring, and management of multiple Spectrum-X switches from a single interface. This is more efficient than manually configuring each switch (A and C) or using SNMP (D), which is primarily for monitoring.

Correct Answer: A

Explanation: Equal-Cost Multi-Path (ECMP) routing is essential for balancing traffic across multiple paths in a network, which is critical for high-performance AI workloads. Static routes (B) do not provide dynamic load balancing. Disabling STP (C) can lead to network loops, and VLAN tagging (D) is unrelated to path load balancing.

Correct Answer: A

Explanation: Creating a separate management VLAN is a best practice for ensuring effective network segmentation and management. This approach isolates administrative traffic from regular data traffic, enhancing security and management efficiency. A flat network topology can lead to scalability issues, while disabling unused ports is a security measure rather than a segmentation strategy. Configuring all ports to maximum speed does not address segmentation or management directly.

Correct Answer: B

Explanation: VXLAN (Virtual Extensible LAN) enables Layer 2 adjacency across Layer 3 boundaries, which is beneficial in virtualized environments where VMs need to communicate as if they are on the same physical network segment. VXLAN does not inherently reduce power consumption, eliminate the need for routers, or encrypt data traffic.

Correct Answer: B

Explanation: Using a pair of switches in an MLAG configuration (B) is a recommended practice for ensuring high availability and redundancy in an AI network. MLAG allows for link aggregation across multiple switches, providing fault tolerance and load balancing. Deploying a single switch (A) does not provide redundancy. Configuring switches in a daisy-chain topology (C) can lead to single points of failure and is not recommended for high availability. Relying solely on software-based failover mechanisms (D) may not provide the necessary hardware redundancy.