Questions & answers (free, no sign-in needed)

10 beginner questions on the Professional Data Engineer exam guide sections: designing data processing systems (data location, IAM), ingesting and processing data (Pub/Sub, Dataflow), storing data (Bigtable, Cloud Storage lifecycle, BigQuery table design), preparing data for analysis (BigQuery ML, materialized views) and maintaining and automating workloads (Airflow orchestration).

Ready to test yourself? Take the graded Beginner test (free; sign in with Google, one click) to get your score.

1. A company collects click events from thousands of mobile apps. It needs a fully managed, scalable service that receives the events asynchronously and decouples the apps from the pipelines that process them. Which service should it use?

   • A. Cloud SQL
   • B. Pub/Sub ✔ correct
   • C. Filestore
   • D. Cloud DNS

   Answer: B. Pub/Sub is an asynchronous, scalable messaging service: publishers send messages to a topic and subscribers (for example a Dataflow pipeline) receive them independently. Cloud SQL is a relational database and Filestore is managed file storage.

   Source: docs.cloud.google.com/pubsub/docs/overview · Domain: Ingesting and processing the data

2. Which Google Cloud service runs Apache Beam pipelines for both batch and streaming data as a fully managed service?

   • A. Datastream
   • B. Cloud Run functions
   • C. Dataflow ✔ correct
   • D. Bigtable

   Answer: C. Dataflow is the managed runner for Apache Beam and handles batch and streaming pipelines with autoscaling. Datastream is a change data capture (CDC) replication service, not a Beam runner.

   Source: docs.cloud.google.com/dataflow/docs/overview · Domain: Ingesting and processing the data

3. An IoT team stores billions of time-series sensor readings and needs low-latency reads and writes by row key at very high throughput. Which storage service fits best?

   • A. Bigtable ✔ correct
   • B. Cloud Storage Archive class
   • C. Memorystore for Redis
   • D. Cloud SQL for MySQL

   Answer: A. Bigtable is a NoSQL wide-column database built for low-latency, high-throughput access to very large data sets keyed by row key, such as time-series and IoT data. Archive storage is for rarely accessed data, and Cloud SQL does not scale this way.

   Source: docs.cloud.google.com/bigtable/docs/overview · Domain: Storing the data

4. A team must automatically delete raw log files from a Cloud Storage bucket 365 days after each file is created. What is the simplest way to do this?

   • A. Write a script that runs every night, lists the bucket and deletes files older than a year
   • B. Turn on Object Versioning for the bucket
   • C. Add a lifecycle rule with a Delete action and an Age condition of 365 days ✔ correct
   • D. Move the bucket to a different region

   Answer: C. Object Lifecycle Management applies rules automatically: a rule with an Age condition and a Delete action removes objects once they reach that age. Object Versioning keeps older copies, so it does the opposite.

   Source: docs.cloud.google.com/storage/docs/lifecycle · Domain: Storing the data

5. Analysts often query a large BigQuery table with filters on event_date and customer_id. Which two table design choices reduce the data that these queries read? (Select TWO)

   • A. Partition the table by event_date ✔ correct
   • B. Cluster the table by customer_id ✔ correct
   • C. Add LIMIT 100 to every query on the non-clustered table
   • D. Grant analysts the BigQuery Data Viewer role
   • E. Copy the table into a second dataset

   Answer: A, B. Partitioning lets BigQuery skip partitions that do not match the date filter, and clustering sorts data so filters on customer_id read fewer blocks. On a non-clustered table, LIMIT does not reduce the bytes read.

   Source: docs.cloud.google.com/bigquery/docs/best-practices-costs · Domain: Storing the data

6. An analyst team knows SQL but not Python. It wants to train a simple forecasting model on data that is already in BigQuery, without moving the data. What should the team use?

   • A. Dataproc with a custom Spark ML job
   • B. Cloud Data Fusion pipelines
   • C. Looker Studio calculated fields
   • D. BigQuery ML with a CREATE MODEL statement ✔ correct

   Answer: D. BigQuery ML lets you create, train and use machine learning models with GoogleSQL statements such as CREATE MODEL and ML.PREDICT, directly where the data lives. The other options need extra tools or do not train models.

   Source: docs.cloud.google.com/bigquery/docs/bqml-introduction · Domain: Preparing and using data for analysis

7. BigQuery materialized views precompute and store the results of a query, which can make repeated dashboard queries faster and cheaper.

   • A. True ✔ correct
   • B. False

   Answer: A. True. A materialized view stores precomputed results and BigQuery keeps them up to date, so queries read the stored results instead of reprocessing the base table each time. This makes them useful for BI dashboards.

   Source: docs.cloud.google.com/bigquery/docs/materialized-views-intro · Domain: Preparing and using data for analysis

8. A company must keep a new BigQuery dataset's data inside the European Union to meet a data residency rule. What should the data engineer do?

   • A. Create the dataset in the US multi-region and add an EU label to it
   • B. Create the dataset in an EU location, which cannot be changed later ✔ correct
   • C. Create the dataset anywhere and move it to the EU at the end of each month
   • D. Store the data in BigQuery and rely on default encryption for residency

   Answer: B. Data residency is controlled by the dataset location, chosen when the dataset is created (for example the EU multi-region or an EU region). After creation the location cannot be changed. Labels and encryption do not control where data is stored.

   Source: docs.cloud.google.com/bigquery/docs/datasets · Domain: Designing data processing systems

9. A team wants a daily workflow that runs a Dataproc job and then loads the results into BigQuery, with task dependencies defined in Python DAGs. Which managed service fits?

   • A. Cloud Composer (Managed Service for Apache Airflow) ✔ correct
   • B. Cloud Scheduler sending one HTTP request
   • C. BigQuery scheduled queries only
   • D. Pub/Sub topics with no subscribers

   Answer: A. Cloud Composer, now named Managed Service for Apache Airflow, is a managed Apache Airflow service. Workflows are Python DAGs that define tasks, their dependencies and schedules. Cloud Scheduler alone triggers a single job; it does not model multi-step pipelines.

   Source: docs.cloud.google.com/composer/docs/composer-3/composer-overview · Domain: Maintaining and automating data workloads

10. When you grant an IAM role on a BigQuery dataset, the role's permissions are inherited by the tables and views inside that dataset.

    • A. True ✔ correct
    • B. False

    Answer: A. True. In BigQuery the project is the parent of datasets, and a dataset is the parent of its tables and views. A role granted at the dataset level is inherited by the resources in it, which helps you design access per dataset.

    Source: docs.cloud.google.com/bigquery/docs/control-access-to-resources-iam · Domain: Designing data processing systems

Original practice questions written by xOps School (DevOpsSchool Private Limited) from the public Google Cloud exam objectives (Professional Data Engineer certification exam guide (current English PDF, checked 2026-10-25)), checked 2026-10-25. They are not actual exam questions. xOps School is not affiliated with or endorsed by Google Cloud; exam names are trademarks of their owners. Official exam page: cloud.google.com/learn/certification/data-engineer.