Google Cloud Platform - Professional Data Engineer

Free GCP - Professional Data Engineer Practice Questions

A handful of real practice questions from our GCP - Professional Data Engineer bank — to give you a true feel for the style and difficulty before you sign up.

1. Which Beam windowing strategy assigns each event to exactly one non-overlapping interval of fixed length?

   • A.Session windows of varying length
   • B.Sliding windows that overlap intervals
   • C.Fixed (tumbling) windows✓ Correct
   • D.Global windows over all events

   Why: Beam windowing primitives differ in how they assign events to intervals. Fixed (tumbling) windows partition the time axis into contiguous non-overlapping intervals of a stated length, so each event maps to exactly one window. The most tempting distractor is sliding windows; sliding windows overlap because their period is shorter than their size, which places each event into multiple windows rather than exactly one as the requirement demands.

2. Analysts need to run SQL over Amazon S3 data without copying the source bytes into Google Cloud. Which BigQuery capability supports this?

   • A.BigQuery federated queries to Cloud SQL
   • B.BigQuery Omni✓ Correct
   • C.BigQuery external tables over Cloud Storage
   • D.BigQuery scheduled queries with an S3 transfer step

   Why: BigQuery Omni runs SQL against data in Amazon S3 or Azure Blob Storage with compute running in the source cloud through Anthos, returning aggregated results to BigQuery without copying the source bytes. Federated queries to Cloud SQL target a different source type and do not reach S3. External tables over Cloud Storage assume the data already resides in Cloud Storage. Scheduled queries with an S3 transfer step explicitly copy the data into BigQuery. BigQuery Omni is the documented cross-cloud query capability for S3 and Azure Blob sources.

3. Planning notes describe a job that reads a fixed Cloud Storage prefix containing the day's Parquet files once and then exits. Which classification fits this source?

   • A.Bounded batch source✓ Correct
   • B.Unbounded streaming source requiring Pub/Sub notifications
   • C.Hybrid source dependent on file watcher cadence
   • D.Continuous source bounded by watermark advancement

   Why: Bounded sources have a known boundary at job start: the job reads the whole input, computes, and exits. A fixed Cloud Storage prefix read once matches that definition. The unbounded streaming option is the most tempting distractor because the same bucket can host an unbounded pattern, but only when a prefix is watched continuously through Pub/Sub notifications rather than read once for a known set of objects.

4. Scheduled loads from Google Ads and YouTube into BigQuery are performed by which managed service?

   • A.Database Migration Service with custom connectors
   • B.Storage Transfer Service over an HTTPS endpoint list
   • C.BigQuery Data Transfer Service✓ Correct
   • D.Datastream with a Pub/Sub destination route

   Why: BigQuery Data Transfer Service handles SaaS connectors such as Google Ads, YouTube, Google Ad Manager, and Search Ads, writing data into BigQuery on a recurring schedule with partition-aware refresh. The most tempting distractor is Datastream, which is a change-data-capture service for relational databases such as MySQL, PostgreSQL, Oracle, and SQL Server and does not connect to SaaS advertising APIs. Storage Transfer Service moves object data, not schema-aware SaaS feeds.

5. Analysts on a logistics carrier's revenue team need to query BigQuery tables that another team has loaded into a shared dataset. Which minimal role pairing lets them execute interactive SQL against those tables?

   • A.roles/bigquery.dataOwner on the dataset.
   • B.roles/storage.objectViewer at the project and roles/bigquery.user on the dataset.
   • C.roles/bigquery.dataViewer on the dataset and roles/bigquery.jobUser on the project.✓ Correct
   • D.roles/bigquery.metadataViewer on the dataset only.

   Why: BigQuery splits the read-data permission from the run-query permission, and a working analyst needs both. dataViewer on the dataset grants table reads, while jobUser at the project supplies the billing context where the job is charged. The closest distractor is granting only dataOwner because it sounds sufficient, but it skips the principle of least privilege and still misses the design intent that the chapter highlights: scope data on the dataset and jobs on the project.

GCP - Professional Data Engineer Exam Details

Number of questions

50

Duration

120 minutes

Passing score

70%

Level

Professional

Delivered by

Google Cloud (GCP)

All figures should be confirmed on the official Google Cloud (GCP) page.

Frequently Asked Questions

Related certifications

• GCP - Associate Data Practitioner
  Google Cloud (GCP)
• AWS SAA-C03
  Amazon Web Services (AWS)
• AZ-104
  Microsoft Azure
• SnowPro Core (COF-C03)
  Snowflake