Google Professional Data Engineer Practice Questions: Picking the Right Service Under Time Pressure
Forty to fifty questions in two hours for $200, and Google publishes no domain weights. The exam has one recurring shape - a scenario with a constraint, and four services that could almost work.
- 40-50Questions
- 120 minTime
- $200Cost
- 2 yearsValid
Table of Contents
The Google Cloud Professional Data Engineer paper is 40-50 questions in 120 minutes and costs $200. Google publishes five sections and no percentages. Treat them as roughly equal, and note that 40-50 questions in 120 minutes gives you around 150 seconds each - enough to read a long scenario properly.
What follows is the question split by domain, then five practice questions written to Google Cloud’s published objectives, each with the reasoning for the right answer and for every wrong one. These are original practice items, not real exam content – the point is to show you how the questions are shaped and where the traps sit.
How the 40-50 questions split
Google Cloud lists the exam sections but does not publish percentage weights for this exam, so nobody can tell you an exact question count per domain. The honest version is below: the sections, and what an even split across a 45-question paper would look like.
| Domain | Weight | If even | |
|---|---|---|---|
| Design data processing systems | not published | ~9 | |
| Ingest and process the data | not published | ~9 | |
| Store the data | not published | ~9 | |
| Prepare and use data for analysis | not published | ~9 | |
| Maintain and automate data workloads | not published | ~9 |
Be sceptical of precise weightings. Sites quoting exact percentages for this exam are inventing them. Google Cloud has not published them, so revise the sections evenly and spend your extra time on whichever one you can least explain out loud.
Five practice questions
Answer each one before reading the key underneath it.
Clickstream events arrive continuously and must be enriched and written to BigQuery within seconds, using the same code as the nightly backfill. Which service fits?
- A Dataproc running a scheduled Spark batch job
- B Dataflow, which runs the same Apache Beam pipeline in streaming and batch
- C Cloud Scheduler triggering a Cloud Function per event
- D BigQuery scheduled queries every five minutes
Answer: B
Dataflow runs Apache Beam pipelines in both streaming and batch modes, which is exactly the requirement to share one codebase between the live path and the backfill.
Why the others fail
A Dataproc is managed Spark and suits lift-and-shift Hadoop work, but a scheduled batch job cannot meet a seconds-level streaming requirement.
C Cloud Scheduler fires on a timetable, not per event, and per-event functions do not share code with a batch pipeline.
D A five-minute schedule cannot deliver seconds-level latency.
An application needs single-digit millisecond reads at very high write throughput for time-series device data, with no need for SQL joins. Which store fits best?
- A Cloud SQL
- B Bigtable
- C BigQuery
- D Firestore
Answer: B
Bigtable is the wide-column store designed for very high throughput and low-latency key-based access, and time-series device data is its canonical use case.
Why the others fail
A Cloud SQL is a managed relational database that does not scale to this write throughput.
C BigQuery is an analytical warehouse. It is superb for scans and aggregation, not for millisecond point reads.
D Firestore is a document database for application state, not high-throughput time-series ingestion.
A BigQuery table is queried almost always by event date, and costs are rising because queries scan the whole table. What is the appropriate change?
- A Cluster the table on a high-cardinality user id
- B Partition the table by event date so queries prune to the dates they need
- C Export the table to Cloud Storage and query it as an external table
- D Create a materialised view of the entire table
Answer: B
Partitioning on the column used for filtering lets BigQuery prune partitions, so a query over three days scans three days rather than the whole table.
Why the others fail
A Clustering helps within partitions and on the columns actually filtered. Clustering on user id does not help a date filter.
C External tables usually make scanning slower and more expensive, not cheaper.
D A materialised view of everything does not reduce the data scanned by a date-filtered query.
A pipeline must not lose records if a downstream system is briefly unavailable. Which design element addresses this directly?
- A Writing directly from the producer to the downstream system with retries
- B Publishing to Pub/Sub, which retains and redelivers messages until acknowledged
- C Increasing the producer's timeout setting
- D Logging failed records to Cloud Logging for later manual replay
Answer: B
Pub/Sub decouples producer from consumer and retains unacknowledged messages for redelivery, so a downstream outage delays processing rather than losing data.
Why the others fail
A Direct writes couple the two systems together; retries help briefly but the producer eventually gives up.
C A longer timeout postpones the failure without providing durability.
D Logs are for diagnosis. Manual replay is not a durability design.
Several pipelines must run in a fixed order with retries and alerting, and the team wants the schedule defined as code. Which service is intended?
- A Cloud Scheduler with one job per pipeline
- B Cloud Composer, managed Apache Airflow, with the DAG defined in code
- C Cron on a Compute Engine instance
- D Dataflow templates triggered manually
Answer: B
Cloud Composer is managed Airflow, where dependencies, retries and alerting are expressed in a DAG defined in Python - precisely the stated requirement.
Why the others fail
A Cloud Scheduler triggers things on a timetable but has no concept of dependency between them.
C A hand-managed cron box reintroduces the operational burden that managed orchestration removes.
D Manual triggering is the opposite of automation.
What catches people out
- Almost every question is service selection. Learn the one-line boundary between Bigtable and BigQuery, Dataflow and Dataproc, Pub/Sub and Cloud Scheduler, and most of the paper resolves.
- The certification lasts two years, shorter than the three years on Google's associate credentials. A renewal exam of 20 questions at $100 exists.
- Google publishes no percentage weights and no pass mark for this exam, so revise the five sections evenly.
- Read the constraint, not the domain. Two options are usually both technically possible and only one meets the latency, cost or throughput limit in the stem.
What it really costs to pass
The sticker price is $200, but that is the cost of passing first time. The number worth budgeting is different.
- First attempt: $200.
- Resit: Google does not discount a resit. Holders inside the renewal window can sit a 20-question renewal exam at $100 rather than the full $200 paper.
- Renewal: the credential lasts 2 years, so the honest cost is the fee divided across that period, plus whatever renewal Google Cloud requires at the end of it.
- Your time: 120 minutes in the chair, and realistically several weeks of preparation before it. That is the largest cost on this list and the only one you cannot pay to avoid.
There is no published pass mark. Google Cloud does not publish one for this exam, so you cannot calculate how many mistakes you can afford. Prepare for a comfortable margin rather than aiming at a threshold nobody has stated.
Who should sit it: there are no formal prerequisites, though Google recommends 3+ years in industry and 1+ year on Google Cloud. Going in well under that bar is usually how people end up paying the fee twice.
How to prepare
A workable sequence for this exam, assuming you already work near the material:
- Read the official guide first. Everything on this page is secondary to what Google Cloud publishes at the link in the sidebar, and guides get revised.
- Work the heaviest domain until you can teach it. Use the table above to decide which one that is.
- Practise under the clock. 120 minutes for 40-50 questions is roughly 160 seconds each. Knowing the material and being able to apply it at that pace are different skills.
- Review every wrong answer to the level of the distractor. Being able to say why the other three options fail is what turns a 60% into a pass.
Frequently asked questions
How many questions are on the Google Professional Data Engineer exam?
Between 40 and 50 multiple-choice and multiple-select questions, with a two-hour limit.
How much does the Professional Data Engineer exam cost?
$200 plus tax. A one-hour, 20-question renewal exam is available at $100 for holders inside the renewal window.
What experience does Google recommend for this exam?
Three or more years of industry experience including at least one year designing and managing solutions on Google Cloud. There are no formal prerequisites.
How long does the Professional Data Engineer certification last?
Two years from the date it is awarded.
Sources
Every figure on this page comes from the certifying body rather than a third-party summary:
- https://cloud.google.com/learn/certification/data-engineer
- https://services.google.com/fh/files/misc/professional_data_engineer_exam_guide_english.pdf
Checked September 2, 2026. Formats and fees change – confirm against Google Cloud before you pay for anything.