01. In Nexus Dashboard Fabric Controller (NDFC), which deployment "Persona" is specifically optimized for managing the leaf-spine fabrics typically used in AI PODs?
a) LAN Fabric
b) IP Fabric for Media
c) SDN Fabric (ACI)
d) SAN Controller
02. An organization is using Cisco Intersight to manage a globally distributed AI infrastructure. Which feature allows them to automate the deployment of OS-level drivers and NVIDIA libraries across all nodes?
a) Intersight Mobile App
b) Intersight Search
c) Intersight Virtual Appliance
d) Intersight Cloud Orchestrator (ICO)
03. When configuring a storage policy for an AI cluster in Cisco Intersight, which setting ensures that the UCS server can utilize NVMe over Fabrics (NVMe-oF) for high-speed data ingestion?
a) Enabling "Local Disk Mirroring"
b) Configuring an "FC-NVMe" or "NVMe over RoCE" interface in the Storage Policy
c) Setting the disk drive speed to 15k RPM
d) Enabling "SATA AHCI" mode in the BIOS
04. You are designing the IP addressing for a large AI cluster using RoCEv2 across a Layer 3 leaf-spine. Why is it recommended to use a small MTU for the Control Plane but a Jumbo MTU (9000+) for the Data Plane?
a) There is no such recommendation; MTU should always be 1500.
b) Because GPUs can only read large packets.
c) Small packets are better for routing updates, while large packets maximize throughput and reduce CPU overhead for the massive AI data transfers.
d) To save money on IP addresses.
05. Why is Redundancy (N+1 or 2N) more complex in AI clusters than in standard web-tier clusters?
a) AI jobs are stateful and long-running; a single node failure can crash a month-long training job.
b) AI nodes don't use IP addresses.
c) GPUs don't support power redundancy.
d) AI traffic is only one-way.
06. A 32-node AI training job has slowed down significantly. An engineer checks the Nexus switch counters and sees a high number of PFC Rx Pause frames on the ports connected to the leaf switches, but zero drops. What is the most likely cause?
a) The fabric is dropping packets due to a hardware failure.
b) The spine switches are congested and are pushing "back-pressure" toward the leaves.
c) The GPUs have stopped sending data.
d) The VLANs are incorrectly configured.
07. In a Hybrid AI deployment, which Cisco solution would be used to provide secure, high-speed connectivity between an on-premises Nexus fabric and a VPC in AWS?
a) Cisco SD-WAN or Cloud Router (CSR/Catalyst 8000V)
b) Cisco Umbrella
c) Cisco Duo
d) Cisco AnyConnect (Standard VPN)
08. Which AI use case is specifically designed to detect anomalies in network traffic patterns to prevent zero-day exploits?
a) Predictive Maintenance
b) Natural Language Processing (NLP)
c) AI-driven Cybersecurity
d) Computer Vision
09. What is the role of a Vector Database in a RAG architecture?
a) It stores the weights of the neural network.
b) It stores high-dimensional "embeddings" of data for fast similarity searches.
c) It acts as the primary DHCP server for the GPUs.
d) It manages the power distribution units (PDUs).
10. During which phase of the AI lifecycle would a data engineer perform "Feature Engineering" and "Data Labeling"?
a) Model Monitoring
b) Data Preparation
c) Model Evaluation
d) Model Deployment
Before you write the Cisco CCNP Data Center (300-640) certification exam, you may have certain doubts in your mind regarding the pattern of the test, the types of questions asked in it, the difficulty level of the questions and time required to complete the questions. These Cisco Certified Specialist - Data Center AI Infrastructure (DCAI) sample questions and demo exam help you in removing these doubts and prepare you to take the test.
