Single-node stress testing and HPL execution: Practice Questions — Cluster Test and Verification (NVIDIA-Certified Professional: AI Infrastructure)
Practice Questions: Single-Node Stress Testing and HPL Execution This set of multiple-choice questions is designed to help candidates prepare for the...
Practice Questions: Single-Node Stress Testing and HPL Execution
This set of multiple-choice questions is designed to help candidates prepare for the Single-Node Stress Testing and HPL Execution portion of the NVIDIA-Certified Professional: AI Infrastructure exam. Each question includes four options, the correct answer, and a brief explanation.
What is the primary purpose of running High Performance Linpack (HPL) during single-node stress testing?
A. To verify network cable signal quality
B. To benchmark floating-point computational performance and stability under load
C. To validate switch firmware versions
D. To test storage throughput
Correct Answer: B
Explanation: HPL is used to measure the floating-point performance of a node and stress the CPU and GPU components to verify stability under heavy computational load.
During single-node stress testing, which of the following is a key indicator that the node is stable?
A. Consistent HPL benchmark results without errors or crashes
B. Successful cable signal verification
C. Firmware update completion
D. High storage I/O rates
Correct Answer: A
Explanation: Stability during stress testing is demonstrated by consistent, error-free HPL runs, indicating the node can handle peak computational workloads reliably.
Which tool is commonly used to initiate and monitor HPL execution on a single node?
A. nvidia-smi
B. mpirun or srun with HPL binary
C. nccopy
D. iperf3
Correct Answer: B
Explanation: HPL is typically launched using MPI execution commands such as mpirun or srun to run the benchmark across available CPU and GPU resources.
What is a common cause of failure during single-node HPL stress testing?
A. Incorrect switch firmware version
B. Insufficient power supply or thermal throttling
C. Cable signal degradation
D. Storage latency issues
Correct Answer: B
Explanation: Power or thermal issues can cause hardware components to throttle or fail, leading to errors or crashes during intensive HPL runs.
Why is it important to run stress tests like HPL on a single node before cluster-wide testing?
A. To verify east-west fabric bandwidth
B. To isolate and confirm node-level hardware and software stability
C. To validate BlueField firmware
D. To test cable signal correctness
Correct Answer: B
Explanation: Single-node stress testing ensures each node is stable and correctly configured before integrating into the larger cluster environment.
Which of the following metrics is NOT typically monitored during HPL execution?
A. Floating-point operations per second (FLOPS)
B. Node temperature and power consumption
C. Network cable signal quality
D. Error rates and system logs
Correct Answer: C
Explanation: Cable signal quality is verified separately; HPL focuses on computational performance and system stability metrics.
What does a sudden drop in HPL performance during a stress test most likely indicate?
A. Firmware is up to date
B. Network fabric bandwidth is sufficient
C. Thermal throttling or hardware instability
D. Successful node assessment by clusterKit
Correct Answer: C
Explanation: A performance drop often signals thermal throttling or hardware issues that reduce computational throughput during stress testing.
Which of the following best describes the role of single-node stress testing in the context of NVIDIA AI infrastructure validation?
A. To confirm cluster-wide fabric bandwidth
B. To verify individual node hardware and software readiness for integration
C. To validate switch and BlueField firmware
D. To perform cable signal quality checks
Correct Answer: B
Explanation: Single-node stress testing is essential to ensure each node is robust and correctly configured before deployment in a multi-node AI infrastructure cluster.