NCA - AI Infrastructure and Operations NCA-AIIO Practice Test 32
Question 310
A GPU operations plan calls for: Distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement. The reason is that a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. Which principle is MOST relevant?
- Virtual GPU technology can expose allocated accelerator capabilities to separate virtual-machine workloads.
- Accelerator virtualization remains constrained by physical memory and supported profile limits.
- Accelerated virtual infrastructure needs supported drivers and entitlements as well as physical GPU capacity.
- Time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG.
Explanation
Correct Answer
D:The governing principle is that time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG. It supports the measure to distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement. because a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. Without the control, a tenant may receive weaker isolation or unpredictable latency than its service contract requires. The most useful acceptance evidence is: The architectural review documents the chosen sharing mode and verifies its actual isolation boundaries.
Incorrect Answers
A:That principle is relevant to this distinct case: A virtual-desktop deployment requires accelerated graphics for several isolated users on shared server GPUs. The associated response would be to evaluate supported NVIDIA vGPU capabilities, profiles and licensing against the application requirements. It does not describe the central reason for this case, which is that time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG.
B:That principle is relevant to this distinct case: A virtualization cluster admits more virtual-GPU memory assignments than can be backed by the available hardware profiles. The associated response would be to plan explicit virtual-GPU profile capacity and admission controls instead of assuming limitless overcommit. It does not describe the central reason for this case, which is that time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG.
C:That principle is relevant to this distinct case: An IT department deploys vGPU workloads but ignores required enterprise license and guest-driver version alignment. The associated response would be to verify supported vGPU software version compatibility and the applicable licensing lifecycle. It does not describe the central reason for this case, which is that time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG.
Question 311
An architecture review proposes: Include interface, RDMA and switch telemetry for link health and congestion diagnosis. The underlying requirement is: A fabric design lists port speeds but omits any means of collecting error and retransmission counters. What technical rationale MOST strongly supports it?
- Priority-based pause settings can introduce broad congestion effects if indiscriminately applied.
- ECN conveys emerging network congestion so senders can reduce demand before severe queue buildup.
- Consistent maximum transmission unit settings matter for efficient and reliable large data transfers.
- Reliable AI network operations need evidence beyond nominal link-up status or configured line rate.
Explanation
Correct Answer
D:The governing principle is that reliable AI network operations need evidence beyond nominal link-up status or configured line rate. It supports the measure to include interface, RDMA and switch telemetry for link health and congestion diagnosis. because a fabric design lists port speeds but omits any means of collecting error and retransmission counters. Without the control, jobs can run slowly for days without a clue that link errors or retransmissions increased. The most useful acceptance evidence is: An injected link impairment produces a correlated alert and counters identifying the affected path.
Incorrect Answers
A:That principle is relevant to this distinct case: A storage and training fabric experiences repeated priority flow-control storms that slow unrelated traffic classes. The associated response would be to inspect PFC pause counters and congestion telemetry before modifying quality-of-service configuration. It does not describe the central reason for this case, which is that reliable AI network operations need evidence beyond nominal link-up status or configured line rate.
B:That principle is relevant to this distinct case: An AI cluster uses ECN-capable switches but network operators disable congestion feedback for RDMA flows. The associated response would be to enable supported explicit congestion notification and endpoint rate-response controls where required. It does not describe the central reason for this case, which is that reliable AI network operations need evidence beyond nominal link-up status or configured line rate.
C:That principle is relevant to this distinct case: A team sets jumbo-frame MTU on GPU servers but overlooks one switch hop in a high-speed data path. The associated response would be to validate path MTU consistency across endpoints, switches and relevant routed interfaces. It does not describe the central reason for this case, which is that reliable AI network operations need evidence beyond nominal link-up status or configured line rate.
Question 312
An architecture review proposes: Distinguish node-local NVLink or NVSwitch scaling from networked multi-node scale-out design. The underlying requirement is: A team adds more GPUs inside one system and assumes the same link topology extends to GPUs in other racks. What technical rationale MOST strongly supports it?
- Adding more distributed nodes expands the operational failure surface as well as compute capacity.
- Useful cluster capacity depends on repair and maintenance reserves as well as nominal GPU count.
- GPU partitioning is a workload-isolation choice, not an automatic substitute for full-device training capacity.
- Scale-up and scale-out use different interconnect domains and communication constraints.
Explanation
Correct Answer
D:The governing principle is that scale-up and scale-out use different interconnect domains and communication constraints. It supports the measure to distinguish node-local NVLink or NVSwitch scaling from networked multi-node scale-out design. because a team adds more GPUs inside one system and assumes the same link topology extends to GPUs in other racks. Without the control, distributed jobs may hit external network limits even when local GPU links have ample bandwidth. The most useful acceptance evidence is: A topology diagram and scale-out test show the actual network path and collective performance.
Incorrect Answers
A:That principle is relevant to this distinct case: A GPU cluster loses progress whenever a worker disappears during a large multi-node training epoch. The associated response would be to plan checkpoint intervals, scheduler restart rules and failure-domain containment for distributed jobs. It does not describe the central reason for this case, which is that scale-up and scale-out use different interconnect domains and communication constraints.
B:That principle is relevant to this distinct case: A research group buys all available GPUs with no spare capacity for maintenance or failed accelerator replacement. The associated response would be to define capacity headroom and controlled drain/replacement procedures based on job availability targets. It does not describe the central reason for this case, which is that scale-up and scale-out use different interconnect domains and communication constraints.
C:That principle is relevant to this distinct case: A multi-GPU training model requires full-GPU memory but a manager proposes partitioning every GPU into tiny MIG slices. The associated response would be to reserve suitable full GPU devices for workloads that need whole-device capacity or communication topology. It does not describe the central reason for this case, which is that scale-up and scale-out use different interconnect domains and communication constraints.
Question 313
The design team faces this case: A text model generates new summaries while a separate classifier assigns each message to a fixed department. It adopts: Distinguish generative modeling from discriminative classification when selecting model evaluation criteria. What evidence BEST validates the intended outcome?
- A simulator run records state transitions, actions and accumulated rewards under the intended policy.
- The evaluation suite scores factual support of summaries separately from the department classification accuracy.
- A review identifies the rule configuration and confirms there is no data-driven parameter optimization step.
- A sample review establishes meaningful cluster properties without claiming the clusters predict confirmed incidents.
Explanation
Correct Answer
B:The strongest proof is: The evaluation suite scores factual support of summaries separately from the department classification accuracy. It directly tests the safeguard to distinguish generative modeling from discriminative classification when selecting model evaluation criteria. The control addresses this specific context: A text model generates new summaries while a separate classifier assigns each message to a fixed department. It matters because generative systems create content or sequences while classifiers predict categories for observed inputs. Without it, testing only category accuracy can miss hallucinated or unsafe content from the generative model.
Incorrect Answers
A:This observation would validate a different control: Recognize reinforcement learning and define rewards, actions and safe simulation boundaries. It fits the case where a robot learns control actions by interacting with a simulated environment and receiving delayed reward signals. It cannot prove success here; instead the required evidence is the evaluation suite scores factual support of summaries separately from the department classification accuracy.
C:This observation would validate a different control: Distinguish fixed rule-based automation from models that learn patterns from examples during training. It fits the case where a project hard-codes hundreds of manually chosen rules to classify support tickets, yet the team describes it as a trained ML model. It cannot prove success here; instead the required evidence is the evaluation suite scores factual support of summaries separately from the department classification accuracy.
D:This observation would validate a different control: Explore unsupervised clustering using selected normalized features and verify cluster interpretability. It fits the case where an operations team has telemetry for thousands of devices but no reliable failure labels and wants to group similar behavior. It cannot prove success here; instead the required evidence is the evaluation suite scores factual support of summaries separately from the department classification accuracy.
Question 314
An architecture review proposes: Inspect CPU NUMA, PCIe and GPU-to-NIC affinity before assigning communication-heavy training jobs. The underlying requirement is: Two GPUs share a host but one is connected through a constrained PCIe path to the preferred network adapter. What technical rationale MOST strongly supports it?
- Optimizer and gradient states can consume substantial memory beyond inference-only model weights.
- GPU provisioning does not replace adequate CPU and system-memory resources for data preparation.
- Reliable training infrastructure must store and restore evolving state as well as original training data.
- PCIe and NUMA topology can affect data paths between CPUs, accelerators and network devices.
Explanation
Correct Answer
D:The governing principle is that pCIe and NUMA topology can affect data paths between CPUs, accelerators and network devices. It supports the measure to inspect CPU NUMA, PCIe and GPU-to-NIC affinity before assigning communication-heavy training jobs. because two GPUs share a host but one is connected through a constrained PCIe path to the preferred network adapter. Without the control, unfavorable placement may add host interconnect traffic and degrade distributed workload throughput. The most useful acceptance evidence is: A topology-aware test reports lower communication latency after matching GPU, CPU and NIC locality.
Incorrect Answers
A:That principle is relevant to this distinct case: A team estimates LLM training memory from the file size of weights and omits optimizer and gradient storage. The associated response would be to model total training-state memory and evaluate supported sharding or memory-saving strategies. It does not describe the central reason for this case, which is that pCIe and NUMA topology can affect data paths between CPUs, accelerators and network devices.
B:That principle is relevant to this distinct case: A GPU server repeatedly runs out of host memory when constructing large mini-batches from many source files. The associated response would be to size host RAM and CPU data-loader workers for preprocessing, buffering and model-feed requirements. It does not describe the central reason for this case, which is that pCIe and NUMA topology can affect data paths between CPUs, accelerators and network devices.
C:That principle is relevant to this distinct case: A training program uses frequent multi-terabyte checkpoints but reserves storage for only its original dataset. The associated response would be to include checkpoint size, frequency, retention and restore throughput in capacity and storage planning. It does not describe the central reason for this case, which is that pCIe and NUMA topology can affect data paths between CPUs, accelerators and network devices.
Question 315
An organization is evaluating an AI deployment: A team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. What solution MOST directly fits the stated need?
- Enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads.
- Plan explicit virtual-GPU profile capacity and admission controls instead of assuming limitless overcommit.
- Distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement.
- Evaluate GPU passthrough or supported direct assignment where workload performance and management justify it.
Explanation
Correct Answer
C:The appropriate response is to distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement. This is specific to the case where a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. The mechanism is that time slicing shares execution opportunity but does not provide the same hardware isolation properties as MIG. If ignored, a tenant may receive weaker isolation or unpredictable latency than its service contract requires. Validate the choice using this targeted evidence: The architectural review documents the chosen sharing mode and verifies its actual isolation boundaries.
Incorrect Answers
A:This alternative addresses a different requirement: An organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. It would be appropriate there to enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads. But the present scenario is about a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. For it, the needed control is to distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement.
B:This alternative addresses a different requirement: A virtualization cluster admits more virtual-GPU memory assignments than can be backed by the available hardware profiles. It would be appropriate there to plan explicit virtual-GPU profile capacity and admission controls instead of assuming limitless overcommit. But the present scenario is about a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. For it, the needed control is to distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement.
D:This alternative addresses a different requirement: A low-latency training appliance requires direct use of a physical GPU, but the deployment plan inserts an unsuitable shared virtual layer. It would be appropriate there to evaluate GPU passthrough or supported direct assignment where workload performance and management justify it. But the present scenario is about a team proposes GPU time slicing as equivalent to hardware partitioning with independent memory fault isolation. For it, the needed control is to distinguish time-shared GPU execution from supported MIG hardware isolation before tenant placement.
Question 316
An operations team implements: Evaluate supported GPUDirect RDMA and the required GPU, NIC, driver and topology compatibility. Which test result would MOST convincingly show the approach works?
- A segmentation test blocks unauthorized tenant-to-management traffic while approved operations continue.
- A validated supported path records lower host-copy overhead with correct end-to-end data delivery.
- End-to-end request tests include external network latency and backend GPU service timing.
- An all-reduce test measures steady collective latency and bandwidth under representative scale.
Explanation
Correct Answer
B:The strongest proof is: A validated supported path records lower host-copy overhead with correct end-to-end data delivery. It directly tests the safeguard to evaluate supported GPUDirect RDMA and the required GPU, NIC, driver and topology compatibility. The control addresses this specific context: A research team expects GPUs to access distributed data without unnecessary staging through host CPU memory. It matters because direct accelerator data paths can reduce host-memory copies when the complete platform supports them. Without it, unsupported assumptions may cause fallback transfers and unexpected CPU overhead.
Incorrect Answers
A:This observation would validate a different control: Segment workload and management paths with required access control and monitored boundaries. It fits the case where a shared GPU cluster combines tenant inference traffic and privileged management access on one unmanaged network. It cannot prove success here; instead the required evidence is a validated supported path records lower host-copy overhead with correct end-to-end data delivery.
C:This observation would validate a different control: Plan client ingress, load balancing and north-south bandwidth separately from training interconnects. It fits the case where a model-serving tier delivers predictions to many remote clients but engineers optimize only GPU-to-GPU links. It cannot prove success here; instead the required evidence is a validated supported path records lower host-copy overhead with correct end-to-end data delivery.
D:This observation would validate a different control: Size east-west network bandwidth and collective latency for the intended distributed training topology. It fits the case where a multi-node transformer training job exchanges large gradients across all GPU workers after each step. It cannot prove success here; instead the required evidence is a validated supported path records lower host-copy overhead with correct end-to-end data delivery.
Question 317
An AI infrastructure team faces this requirement: An organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. Which action BEST addresses the operational constraint?
- Assess GPU Multi-Instance GPU partitions on supported hardware with compatible workload sizes.
- Enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads.
- Verify supported vGPU software version compatibility and the applicable licensing lifecycle.
- Plan explicit virtual-GPU profile capacity and admission controls instead of assuming limitless overcommit.
Explanation
Correct Answer
B:The appropriate response is to enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads. This is specific to the case where an organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. The mechanism is that gPU virtualization requires isolation at management, device allocation and data access layers. If ignored, a misconfigured guest may gain access to shared device resources or another tenant's data. Validate the choice using this targeted evidence: An unauthorized guest test cannot view unassigned device resources or protected host management endpoints.
Incorrect Answers
A:This alternative addresses a different requirement: A trusted internal platform needs to run several smaller inference services on one MIG-capable GPU with reserved resources for each workload. It would be appropriate there to assess GPU Multi-Instance GPU partitions on supported hardware with compatible workload sizes. But the present scenario is about an organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. For it, the needed control is to enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads.
C:This alternative addresses a different requirement: An IT department deploys vGPU workloads but ignores required enterprise license and guest-driver version alignment. It would be appropriate there to verify supported vGPU software version compatibility and the applicable licensing lifecycle. But the present scenario is about an organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. For it, the needed control is to enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads.
D:This alternative addresses a different requirement: A virtualization cluster admits more virtual-GPU memory assignments than can be backed by the available hardware profiles. It would be appropriate there to plan explicit virtual-GPU profile capacity and admission controls instead of assuming limitless overcommit. But the present scenario is about an organization consolidates multiple AI tenants but has no process to constrain access to virtual GPU devices and images. For it, the needed control is to enforce hypervisor, guest-driver and tenant resource permissions for virtualized accelerated workloads.
Question 318
A data-center manager declines to validate approved facility life-safety procedures and equipment integration with qualified engineers. The operational context is: Facility operations adds high-density racks without reviewing fire detection and emergency shutoff responsibilities. Which risk is MOST direct?
- A leak or humidity excursion may damage systems before a routine service inspection.
- Damaged fiber or unlabeled cables can create intermittent link loss and lengthy repair outages.
- Inadequate emergency procedures can expose people and equipment to preventable hazards.
- A short upstream feed interruption can stop every accelerator node in the hall.
Explanation
Correct Answer
C:The primary exposure is that inadequate emergency procedures can expose people and equipment to preventable hazards. This follows from facility operations adds high-density racks without reviewing fire detection and emergency shutoff responsibilities. where the omitted measure was to validate approved facility life-safety procedures and equipment integration with qualified engineers. The technical reason is that fire detection and safety systems must be planned alongside accelerator and cooling deployment. A good operational validation after repair would be: The facility sign-off confirms appropriate detection, response ownership and approved emergency controls.
Incorrect Answers
A:This exposure arises in a different situation: A new AI room has reliable cooling capacity but no monitoring for localized humidity or leak alarms. There, the corresponding safeguard is to add appropriate environmental sensors and integrate alerts with owned operational response procedures. In the present case, the more direct risk is that inadequate emergency procedures can expose people and equipment to preventable hazards.
B:This exposure arises in a different situation: A new cluster has redundant high-speed links but cable trays cannot support the planned fiber and copper bundles. There, the corresponding safeguard is to design documented cable routes, bend radii and labeling for power and network infrastructure. In the present case, the more direct risk is that inadequate emergency procedures can expose people and equipment to preventable hazards.
D:This exposure arises in a different situation: A GPU hall relies on one unprotected utility feed although the business requires continued operation through feed faults. There, the corresponding safeguard is to assess electrical redundancy, UPS autonomy and generator integration against the site's availability objective. In the present case, the more direct risk is that inadequate emergency procedures can expose people and equipment to preventable hazards.
Question 319
A cluster engineer is diagnosing the following situation: A GPU cluster has healthy operating-system pings but the platform team cannot see accelerator or board-level fault events. Which early operational signal MOST directly points to the relevant infrastructure issue?
- Automated remediation is scheduled against healthy GPUs unrelated to the detected fault source.
- Operators repeatedly restart jobs on one device without attaching a shared hardware incident record.
- A sudden cluster-wide failure follows simultaneous fabric and firmware changes with no separate change trail.
- BMC records a GPU hardware fault while normal host connectivity and CPU checks remain healthy.
Explanation
Correct Answer
D:The relevant warning signal is: BMC records a GPU hardware fault while normal host connectivity and CPU checks remain healthy. That observation reflects the situation in which a GPU cluster has healthy operating-system pings but the platform team cannot see accelerator or board-level fault events. It motivates the response to integrate supported BMC and accelerator telemetry into the central monitoring and incident workflow. The distinction matters because node availability cannot be inferred from network reachability alone when hardware components can degrade independently. Confirm resolution through this independent check: An authorized simulated hardware-health event identifies the affected node and reaches the correct on-call owner.
Incorrect Answers
A:That sign is most pertinent to another operational case: Engineers trigger automated node remediation from broad alerts without checking device identities or active jobs. For that case the appropriate measure is to scope remediation to verified affected nodes and require approval for destructive maintenance actions. Here the decisive symptom is bMC records a GPU hardware fault while normal host connectivity and CPU checks remain healthy.
B:That sign is most pertinent to another operational case: A node reports intermittent GPU failures but operators simply restart the job without recording the incident context. For that case the appropriate measure is to create an owned incident workflow with timeline, affected device identity, evidence and escalation rules. Here the decisive symptom is bMC records a GPU hardware fault while normal host connectivity and CPU checks remain healthy.
C:That sign is most pertinent to another operational case: A team changes cluster networking and driver versions on the same day without an approval window or recovery plan. For that case the appropriate measure is to use governed change records, dependency checks and validated rollback procedures for cluster-wide updates. Here the decisive symptom is bMC records a GPU hardware fault while normal host connectivity and CPU checks remain healthy.
Question 320
An organization is evaluating an AI deployment: A data center plans more training workers despite existing GPUs regularly hitting enforced power limits. What solution MOST directly fits the stated need?
- Monitor board power draw, configured limits and resulting effective clock rates before expanding hardware.
- Correlate temperature, clock state and thermal-throttling reasons during sustained workload measurements.
- Measure host-device transfer duration, PCIe utilization and data-path topology during the real workload.
- Use NVIDIA DCGM monitoring and supported telemetry exporters for centralized device diagnostics.
Explanation
Correct Answer
A:The appropriate response is to monitor board power draw, configured limits and resulting effective clock rates before expanding hardware. This is specific to the case where a data center plans more training workers despite existing GPUs regularly hitting enforced power limits. The mechanism is that power-cap telemetry helps distinguish policy-limited compute throughput from ordinary algorithm inefficiency. If ignored, adding nodes may repeat the same underperformance if electrical limits remain unaddressed. Validate the choice using this targeted evidence: A controlled comparison reports stable power and throughput at a facilities-approved operating limit.
Incorrect Answers
B:This alternative addresses a different requirement: An accelerator runs at reduced clock speed during a full training epoch but device utilization appears high. It would be appropriate there to correlate temperature, clock state and thermal-throttling reasons during sustained workload measurements. But the present scenario is about a data center plans more training workers despite existing GPUs regularly hitting enforced power limits. For it, the needed control is to monitor board power draw, configured limits and resulting effective clock rates before expanding hardware.
C:This alternative addresses a different requirement: GPU utilization looks normal but a pipeline spends much of each step copying tensors through a constrained PCIe link. It would be appropriate there to measure host-device transfer duration, PCIe utilization and data-path topology during the real workload. But the present scenario is about a data center plans more training workers despite existing GPUs regularly hitting enforced power limits. For it, the needed control is to monitor board power draw, configured limits and resulting effective clock rates before expanding hardware.
D:This alternative addresses a different requirement: A fleet has thousands of accelerators but operators manually run nvidia-smi on each node to identify unhealthy GPUs. It would be appropriate there to use NVIDIA DCGM monitoring and supported telemetry exporters for centralized device diagnostics. But the present scenario is about a data center plans more training workers despite existing GPUs regularly hitting enforced power limits. For it, the needed control is to monitor board power draw, configured limits and resulting effective clock rates before expanding hardware.