NCP-GENL Production Monitoring and Reliability Practice Question
An LLM inference service deployed on NVIDIA Triton Inference Server is experiencing occasional failures under high load. The team wants to implement proactive monitoring to predict and prevent these failures. Which two metrics should be prioritized for early detection of potential issues? (Choose two.)
⚠ Common exam trap
The trap here is focusing on general system metrics like CPU or disk I/O, which are less relevant for GPU-bound LLM inference under load.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
GPU memory utilization per instance
Under high load, LLM inference failures often stem from resource exhaustion or overload. GPU memory utilization is critical because insufficient memory leads to out-of-memory errors. Request queue time indicates if the system is falling behind, predicting timeouts. Together, these metrics provide early warning of impending failures, enabling proactive measures.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Disk I/O throughput
Why it's wrong here
Disk I/O is generally not a bottleneck for LLM inference once models are loaded into GPU memory. High disk I/O might affect model loading but not steady-state inference. Therefore, it is not a priority for early detection of failures during high load. Other metrics like GPU memory and queue time are more relevant.
- ✗
CPU utilization of the Triton server process
Why it's wrong here
CPU utilization is less critical for GPU-accelerated LLM inference, as the heavy computation occurs on GPUs. While CPU can be a bottleneck for preprocessing, it is typically not the primary cause of failures under high load. Monitoring it may be useful but not as predictive as GPU memory or request queue time.
- ✗
Number of active connections
Why it's wrong here
While active connections indicate load, they do not directly predict failures. A high number of connections may be normal if the server can handle them. Without context on queue times or resource usage, this metric alone is less actionable. It is not as directly tied to imminent failures as memory or queue metrics.
- ✓
GPU memory utilization per instance
Why this is correct
GPU memory utilization is critical because LLMs are memory-intensive. Rising memory usage can indicate memory leaks or increased batch sizes, leading to out-of-memory errors. Monitoring this metric allows proactive scaling or optimization before failures occur. It directly relates to resource exhaustion, a common cause of inference failures under load.
- ✓
Request queue time
Why this is correct
Request queue time reflects how long requests wait before being processed. Increasing queue times indicate that the server is struggling to keep up with demand, which can lead to timeouts and failures. Monitoring queue time helps predict overload conditions and allows for timely intervention, such as scaling or load shedding.
About these practice questions
One of 352 original NCP-GENL practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official NVIDIA exam blueprint
This NCP-GENL practice question is part of Courseiva's free NVIDIA certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the NCP-GENL exam.