The answer is to use a machine type with more CPU cores to parallelize embedding lookups. This is correct because when troubleshooting inference latency due to a CPU bottleneck, the key indicator is 100% CPU utilization alongside low GPU usage—the CPU cannot feed the GPU fast enough, so adding more CPU cores directly addresses the serialized embedding table lookups that are causing the linear latency increase with batch size. On the Google Professional Machine Learning Engineer exam, this scenario tests your ability to diagnose resource contention in real-time serving, a common trap where candidates mistakenly scale replicas or upgrade GPUs without first identifying the actual bottleneck. The memory tip here is “follow the bottleneck”—always check utilization metrics before scaling; if CPU is pegged and GPU is idle, more cores, not more GPUs, will reduce latency.
PMLE Serving and scaling models Practice Question
This PMLE practice question tests your understanding of serving and scaling models. Read the scenario carefully and evaluate each option against the stated constraints before committing to an answer. After answering, compare your reasoning against the explanation and wrong-answer breakdown below. Once you have made your selection, read the full explanation to reinforce the concept and understand why each distractor is designed to mislead on exam day.
You are responsible for deploying a real-time recommendation model that uses a large embedding table (5 GB) and a small neural network. The model is served through a custom container on Vertex AI Prediction. The end-to-end latency requirement is under 200 ms. During load testing with 500 QPS, you observe that latency increases linearly with batch size. You are currently using a single replica with an n1-standard-8 machine and one T4 GPU. The embedding table is loaded entirely in GPU memory. However, CPU utilization is at 100% while GPU is at 30%. What is the best approach to meet the latency requirement at scale?
Clue words in this question
Noticing these words before you look at the options changes how you read each choice.
Clue: "best"
Why it matters: Signals that multiple options may be partially correct. Choose the option that most directly solves the exact problem described, not the one that sounds most complete.
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Use a machine type with more CPU cores to parallelize embedding lookups.
Option D is correct because CPU is the bottleneck; using a machine type with more CPU cores (e.g., n1-highcpu-16) allows parallel embedding lookups and reduces latency. Option A increases resources but not in the bottleneck area. Option B increases replicas but each would still be CPU-bound. Option C is expensive and may not improve latency if model not T PU-compatible.
Key principle: OSPF neighbour adjacency depends on matching area, hello/dead timers, network type, and authentication — IP reachability alone is not enough.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
✗
Increase the number of replicas and use a global load balancer to distribute traffic.
Why it's wrong here
Each replica still has CPU bottleneck; adding replicas increases cost but not per-request latency.
✗
Use a custom container that partitions the embedding table across multiple GPUs within a single replica.
Why it's wrong here
GPU is underutilized; CPU bottleneck remains.
✗
Switch to a TPU v2-8 pod slice to accelerate embedding lookups.
Why it's wrong here
TPUs are designed for large models; may not help with CPU-bound embedding lookups.
✓
Use a machine type with more CPU cores to parallelize embedding lookups.
Why this is correct
More CPU cores reduce contention and latency for embedding operations.
Clue confirmation
The clue word "best" in the question point toward this answer.
Common exam trap: OSPF can fail even when IP connectivity looks correct
OSPF neighbour formation depends on matching areas, timers, network type, authentication and passive-interface behaviour. Do not choose an answer only because the devices can ping.
Detailed technical explanation
How to think about this question
OSPF questions usually test the details that control adjacency and route selection. Read the neighbour state, area, router ID and interface configuration before deciding what is wrong.
KKey Concepts to Remember
OSPF neighbours must agree on key parameters.
Router ID selection can affect neighbour relationships and LSDB output.
OSPF cost influences the preferred path.
A route can appear in OSPF information but not become the installed route.
TExam Day Tips
→Check area mismatch first when OSPF adjacency fails.
→Review passive interfaces when a network is advertised but no neighbour forms.
→Use show ip ospf neighbor and show ip route clues carefully.
Key takeaway
OSPF neighbour adjacency depends on matching area, hello/dead timers, network type, and authentication — IP reachability alone is not enough.
Real-world example
How this comes up in practice
An e-commerce site experiences heavy traffic on Black Friday and near-zero traffic during off-peak weeks. Rather than provisioning permanent large VMs, the team uses auto-scaling groups that add capacity automatically under load and reduce it overnight. Questions like this test whether you understand elasticity, availability zones, and cloud compute scaling patterns.
What to study next
Got this wrong? Here's your next step.
Review OSPF neighbour requirements — matching area type, hello and dead timers, network type, stub flags, and authentication. Study show ip ospf neighbor states (INIT, 2-WAY, FULL). Then practise related PMLE OSPF questions on adjacency and route selection.
Serving and scaling models — This question tests Serving and scaling models — OSPF neighbours must agree on key parameters..
What is the correct answer to this question?
The correct answer is: Use a machine type with more CPU cores to parallelize embedding lookups. — Option D is correct because CPU is the bottleneck; using a machine type with more CPU cores (e.g., n1-highcpu-16) allows parallel embedding lookups and reduces latency. Option A increases resources but not in the bottleneck area. Option B increases replicas but each would still be CPU-bound. Option C is expensive and may not improve latency if model not T PU-compatible.
What should I do if I get this PMLE question wrong?
Review OSPF neighbour requirements — matching area type, hello and dead timers, network type, stub flags, and authentication. Study show ip ospf neighbor states (INIT, 2-WAY, FULL). Then practise related PMLE OSPF questions on adjacency and route selection.
Are there clue words in this question I should notice?
Yes — watch for: "best". Signals that multiple options may be partially correct. Choose the option that most directly solves the exact problem described, not the one that sounds most complete.
What is the key concept behind this question?
OSPF neighbours must agree on key parameters.
About these practice questions
Courseiva creates original exam-style practice questions with explanations and wrong-answer analysis. It does not publish real exam questions, exam dumps, or protected exam content. Learn why practice questions differ from exam dumps →
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
This PMLE practice question is part of Courseiva's free Google Cloud 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 PMLE exam.
Question Discussion
Share a tip, memory trick, or ask about the reasoning behind this question. Do not post real exam questions, leaked content, braindumps, or copyrighted exam material. Comments are moderated and may be removed without notice.
Sign in to join the discussion.