Free Designing Cisco Data Center Infrastructure Practice Questions

Sample Practice Questions

Try these 10 sample questions from the Designing Cisco Data Center Infrastructure practice bank. Each includes the answer and explanation. Continue with the interactive practice above to review the full question set.

1. 1A deployed image-recognition service applies an already trained neural network to newly uploaded images without updating its learned parameters. Which AI activity is it performing?

   1. A. Training a model from an initial dataset
   2. B. Performing inference on new input data
   3. C. Fine-tuning the model using new labels
   4. D. Retraining the model after concept drift

   Answer: B

   Explanation: Cisco states, "Inference frameworks take the knowledge from trained neural network models and apply them to new data to predict the outcomes." The service uses learned knowledge on new images; training, fine-tuning, and retraining would update the model.

   Reference: Cisco Data Center Networking Blueprint for AI/ML Applications — Introduction

2. 2A UCS fabric interconnect runs Ethernet end-host mode. Which forwarding restriction prevents it from acting as a transit bridge between its upstream Ethernet ports?

   1. A. It blocks all server-to-uplink traffic until spanning tree converges
   2. B. It denies forwarding from one uplink port to another uplink port
   3. C. It forwards each server frame simultaneously through every uplink
   4. D. It uses STP to select one forwarding uplink for the entire domain

   Answer: B

   Explanation: Cisco states that end-host mode avoids loops by "denying uplink ports from forwarding traffic to each other". It does not run STP and pins server traffic rather than flooding egress traffic over every uplink.

   Reference: Cisco UCS Manager Network Management Guide Using the CLI, Release 4.2 — LAN Connectivity — Ethernet Switching Mode > End-Host Mode

3. 3An ONTAP iSCSI deployment with ESXi must tolerate loss of one physical storage-network switch. Which host-connectivity plan provides the required independent network paths?

   1. A. Two VMkernel addresses using one physical NIC and one storage switch
   2. B. Multiple physical NICs connected through multiple physical switches
   3. C. Two IP aliases on a single physical storage interface and cable
   4. D. Two target names reached through one host NIC and one storage switch

   Answer: B

   Explanation: NetApp states, "You can also use multiple physical NICs connected to multiple physical switches to provide HA and increased throughput." This removes the single-switch dependency. Extra logical addresses or target names on one physical path do not survive failure of its only switch.

   Reference: SAN (FC, FCoE, NVMe/FC, iSCSI), RDM — ONTAP and VMware vSphere — Best practices > IP-based SAN multipath connectivity

4. 4An external application will execute supported Nexus CLI commands through NX-API CLI and consume structured responses. Which request design matches the documented interface?

   1. A. Use an HTTP/HTTPS GET with CLI commands in the URI query
   2. B. Use an HTTP/HTTPS PUT with CLI commands in the request body
   3. C. Use an HTTP/HTTPS POST with CLI commands in the request body
   4. D. Use an HTTP/HTTPS DELETE with CLI commands in the request body

   Answer: C

   Explanation: Cisco states, "NX-API uses HTTP/HTTPS as its transport. CLIs are encoded into the HTTP/HTTPS POST body." NX-API CLI therefore uses POST body encoding and can return XML or JSON. GET query placement, PUT, and DELETE are different HTTP request designs rather than this documented CLI request interface.

   Reference: Cisco Nexus 9000 Series NX-OS Programmability Guide, Release 10.4(x) — NX-API CLI — About NX-API CLI > Transport; Using NX-API CLI

5. 5Which characteristic of deep learning explains its ability to build models directly from raw input data?

   1. A. All useful input representations are specified manually before learning
   2. B. Feature extraction is kept separate from learning in the neural network
   3. C. Raw examples are matched against an unchanged set of handwritten rules
   4. D. Adaptable neural layers learn useful representations from raw examples

   Answer: D

   Explanation: Cisco states, "Deep learning models have highly flexible architectures that allow them to learn directly from raw data." Learned representations within adaptable neural layers explain this capability. Mandatory manual features, separate fixed feature extraction, and unchanged handwritten rules instead describe misconceptions about how deep learning learns from raw examples.

   Reference: Cisco Data Center Networking Blueprint for AI/ML Applications — Introduction

6. 6A vNIC in UCS Ethernet end-host mode is hard pinned to a single uplink. There is no fabric failover for this vNIC. What happens if that uplink fails?

   1. A. The vNIC automatically spreads its traffic across all remaining uplinks
   2. B. The FI changes to Ethernet switch mode to preserve the vNIC's traffic
   3. C. The vNIC moves to the other fabric without requiring a failover setting
   4. D. The vNIC remains down because the system cannot dynamically repin it

   Answer: D

   Explanation: Cisco states, "if a vNIC is hard pinned to an uplink port and this uplink port goes down, the system cannot repin the vNIC, and the vNIC remains down." Hard pinning is not dynamic uplink selection, and the scenario provides no fabric-failover path.

   Reference: Cisco UCS Manager Network Management Guide Using the CLI, Release 4.2 — LAN Connectivity — Ethernet Switching Mode > End-Host Mode > Note

7. 7A designer is following NetApp's ONTAP/vSphere guidance for iSCSI with multiple virtual switches. Which VMkernel addressing approach does that guidance recommend?

   1. A. One VMkernel interface on a routed management subnet with NIC teaming
   2. B. Multiple VMkernel interfaces on one routed storage subnet with NIC teaming
   3. C. Multiple VMkernel interfaces on different unrouted storage subnets with NIC teaming
   4. D. Multiple VMkernel interfaces on one unrouted storage subnet with port binding

   Answer: C

   Explanation: NetApp states, "For iSCSI or NVMe/TCP networks, use multiple VMkernel network interfaces on different unrouted network subnets with NIC teaming when multiple virtual switches are present." That specific multiple-vSwitch recommendation selects different unrouted storage subnets. Other layouts can have different uses, but are not the recommendation asked for here.

   Reference: SAN (FC, FCoE, NVMe/FC, iSCSI), RDM — ONTAP and VMware vSphere — Best practices > IP-based SAN networks

8. 8In model-driven NX-OS automation, which language describes the structure of configuration and operational data, RPCs, and notifications, rather than merely encoding a message payload?

   1. A. YANG
   2. B. JSON
   3. C. XML
   4. D. YAML

   Answer: A

   Explanation: Cisco states, "YANG is a data modeling language used to describe configuration and operational data, remote procedure calls, and notifications for network devices." JSON, XML, and YAML can represent serialized data; they do not replace the YANG model defining this device interface.

   Reference: Cisco Nexus 9000 Series NX-OS Programmability Guide, Release 10.4(x) — Overview — About Model-Driven Programmability

9. 9What role does a DPU incorporated into a SmartNIC provide in an AI data center?

   1. A. Offloading network, security, and storage data processing from host CPUs
   2. B. Executing general-purpose application code as the host's primary processor
   3. C. Performing the model's tensor computations as its primary GPU accelerator
   4. D. Providing link connectivity while leaving all data-path processing on the host

   Answer: A

   Explanation: NVIDIA describes "A rich set of flexible and programmable acceleration engines that offload and improve applications performance for AI and machine learning, zero-trust security, telecommunications and storage, among others." The DPU provides infrastructure/data-path offload. The alternatives confuse this role with the host CPU, the model-compute GPU, or a connectivity-only NIC.

   Reference: What Is a DPU? — CPU v GPU v DPU: What Makes a DPU Different?

10. 10A UCS design explicitly requires the fabric interconnect to participate in spanning tree as a traditional Ethernet switch. Which Ethernet mode provides that behavior?

    1. A. Switch mode, which runs STP to avoid loops
    2. B. End-host mode, which runs STP on every uplink
    3. C. End-host mode, which enables STP after hard pinning
    4. D. Switch mode, which removes STP through vNIC pinning

    Answer: A

    Explanation: Cisco states, "The fabric interconnect runs STP to avoid loops" under Switch Mode. End-host mode uses pinning and uplink forwarding restrictions rather than STP. Switch mode does not remove STP through pinning.

    Reference: Cisco UCS Manager Network Management Guide Using the CLI, Release 4.2 — LAN Connectivity — Ethernet Switching Mode > Switch Mode