hardMultiple Select
PDE Practice Question: Which THREE considerations are important when…
Which THREE considerations are important when designing a batch prediction pipeline for a large dataset on Vertex AI?
⚠ Common exam trap
Google Cloud often tests the misconception that batch prediction requires a real-time endpoint or automatically uses GPUs, when in fact batch prediction is a serverless, endpoint-free process that requires explicit machine type and GPU configuration.
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
✓
Choosing the appropriate machine type (e.g., n1-standard-16) balances cost and throughput
Option C is correct because the machine type selected for a Vertex AI batch prediction job (for example n1-standard-16) directly determines the CPU, memory, and accelerator resources available, so it must be sized to balance cost against throughput for a large dataset. Option D is correct because Vertex AI batch prediction shards input across worker nodes, and splitting very large input files into multiple smaller files (e.g., many shards in Cloud Storage) increases parallelism and reduces per-file processing bottlenecks. Option E is correct because batch prediction reads input from Cloud Storage and requires a supported format such as JSONL, CSV, or TFRecord with the appropriate instance keys, so the data must be staged there in a compatible layout. Option A is not correct because batch prediction does not automatically attach GPUs based on the model framework; you must explicitly configure accelerators, and many frameworks run fine on CPU. Option B is not correct because batch prediction is an offline, asynchronous job that does not require or use a dedicated real-time endpoint, which is only needed for online prediction.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Batch prediction automatically uses GPUs if the model framework requires them
Why it's wrong here
Vertex AI batch prediction uses the machine type and accelerators you explicitly configure in the job's compute resources; it does not automatically attach GPUs based on framework. Automatic accelerator selection is tempting because managed training can infer hardware, but batch prediction leaves accelerator choice to the submitted configuration.
- ✗
Batch prediction requires a dedicated real-time endpoint
Why it's wrong here
Batch prediction jobs run asynchronously against input data and write output to a destination such as BigQuery or Cloud Storage; they do not require a deployed real-time endpoint. A dedicated endpoint is tempting because it is needed for online prediction, where low-latency per-request serving is the requirement.
- ✓
Choosing the appropriate machine type (e.g., n1-standard-16) balances cost and throughput
Why this is correct
Batch prediction throughput and cost scale with the machine type chosen for the prediction nodes; larger types process more rows concurrently but bill higher. Selecting an appropriate machine type therefore directly satisfies the stem's requirement to balance cost against throughput for a large dataset.
- ✓
Large input files can be split into multiple smaller files to improve parallelism
Why this is correct
Vertex AI batch prediction shards input across worker nodes; splitting large files into smaller ones increases the number of shards, raising parallelism and reducing overall job duration. This directly satisfies the stem's requirement to improve parallelism when designing a pipeline for a large dataset.
- ✓
Input data should be in Cloud Storage in a format supported by Vertex AI (e.g., JSONL, TFRecord)
Why this is correct
Vertex AI batch prediction reads input directly from Cloud Storage and requires supported formats such as JSONL or TFRecord; unsupported or malformed data causes the job to fail. Storing input in Cloud Storage in a supported format therefore satisfies the stem's data-format requirement for a large batch pipeline.
Go deeper
Related to this question
About these practice questions
One of 747 original PDE 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 by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This PDE 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 PDE exam.