Professional Data Engineer

Last checked October 22, 2026 against the official Google Cloud exam page.

1. Question 1Design data processing systems

   A bank must process card transactions in near real time. Its requirement is that a duplicate publish from the upstream system never creates two ledger entries. Dataflow will write to BigQuery. Which approach best meets this?

   1. AWrite every event to Cloud Storage first and run a nightly batch job that overwrites the ledger table
   2. BRely on Pub/Sub's at-least-once delivery and deduplicate with a unique transaction ID in the Dataflow pipeline before writing
   3. CSet the Dataflow sink to streaming inserts with insertId derived from the Pub/Sub message ID and assume that eliminates all duplicates forever
   4. DUse Pub/Sub exactly-once delivery alone and skip any deduplication step in the pipeline
   Show answer and explanationHide answer

   Correct answer: B

   Pub/Sub delivers at least once, so the pipeline must tolerate repeats, and deduplicating on a business key such as the transaction ID is the reliable pattern. Exactly-once delivery is a per-subscription option with conditions and does not cover every upstream retry. BigQuery insertId deduplication is best effort within a short window, so it cannot be relied on as a permanent guarantee.

2. Question 2Store the data

   A smart-meter platform writes several hundred thousand readings per second. Engineers need single-digit millisecond lookups of the latest readings for one meter ID and range scans over time for that meter. Which storage choice and key design best fit?

   1. ACloud Spanner with a primary key that starts with a monotonically increasing timestamp
   2. BBigQuery streaming inserts into a table partitioned by ingestion time and clustered by meter ID
   3. CCloud Bigtable with a row key that starts with a reversed meter ID followed by a timestamp
   4. DCloud SQL for PostgreSQL with read replicas and a B-tree index on meter ID and timestamp
   Show answer and explanationHide answer

   Correct answer: C

   Bigtable handles very high write throughput with low-latency key-based reads and range scans, and a reversed or hashed prefix spreads writes so they do not hotspot. BigQuery is an analytical store and is not built for millisecond point reads. A timestamp-first key in Spanner would concentrate writes on one split.

3. Question 3Maintain and automate data workloads

   A Dataflow streaming job that reads from Pub/Sub and writes to BigQuery must be updated to add a new transform without losing its place in the stream. Which action should the engineer take?

   1. ACreate a second job with the same name and let both write to the table until the old one is deleted
   2. BUpdate the running job in place, which preserves its pipeline state and keeps the same job identity
   3. CDrain the job, which stops reading and flushes in-flight work, then start the new version from the Pub/Sub snapshot taken earlier
   4. DCancel the job, which discards in-flight elements, and start a new job from the latest offset
   Show answer and explanationHide answer

   Correct answer: B

   Streaming Dataflow jobs can be updated in place when the change is compatible, which preserves the job and its state. Cancelling loses in-flight work, and draining is for stopping a job cleanly rather than preserving the same job through a compatible update. Running two jobs writing to one table risks duplicates and inconsistent output.

4. Question 4Ingest and process data

   A retailer streams clickstream events to Pub/Sub from mobile apps. Events from users on poor connections can arrive up to 10 minutes after they happened. The team needs hourly counts of sessions based on when the user acted, not when the event reached the pipeline. Which Dataflow design fits this requirement?

   1. AFixed one-hour windows on processing time with zero allowed lateness
   2. BFixed one-hour event-time windows with a watermark-based trigger and 15 minutes of allowed lateness
   3. CSession windows built from Pub/Sub publish timestamps with no triggers
   4. DA global window with a count-based trigger every 1,000 elements
   Show answer and explanationHide answer

   Correct answer: B

   Event-time windows assign each event to the hour the user acted, and allowed lateness keeps late events that arrive within the 10-minute delay. Processing-time windows would put late events into the wrong hour, and publish time is not the user action time. Session windows answer a different question from hourly counts.

5. Question 5Prepare and use data for analysis

   An analytics team wants analysts to reuse a set of transformations on raw BigQuery tables, with version control, dependency ordering and tests, all managed from a Git repository. Which Google Cloud capability fits this need?

   1. ADatastream
   2. BCloud Data Fusion
   3. CLooker Studio scheduled extracts
   4. DDataform
   Show answer and explanationHide answer

   Correct answer: D

   Dataform compiles SQL transformation workflows for BigQuery with dependency graphs, assertions and Git integration. Cloud Data Fusion is an ETL integration service aimed at pipeline building rather than SQL transformation projects. Looker Studio is a visualization tool, and Datastream replicates change data from databases.