Professional Data Engineer
Free Google Cloud Certified - Professional Data Engineer Practice Test
12 exam-style questions with full explanations — no sign-up. Score yourself, then close your gaps with the full course.
Exam fee ~$2005 exam domainsLevel Advanced2 timed practice tests in the course
✅ Free practice — no sign-up📝 Real exam-style questions💡 Detailed explanations💸 30-day money-back via Udemy
A logistics company ingests telemetry from 2 million IoT sensors, sustaining 500,000 writes per second. Application dashboards require single-digit millisecond p99 read latency on individual sensor keys, and the dataset will grow to multiple petabytes with a flexible wide-column schema. Which storage service BEST meets these requirements?
Bigtable is a wide-column, key-value NoSQL database purpose-built for very high-throughput writes and single-digit millisecond latency at petabyte scale, which matches the sensor telemetry workload exactly.
A retail company streams point-of-sale transactions through Pub/Sub into a Dataflow pipeline using fixed 5-minute event-time windows. Store network outages cause some transaction events to arrive up to 10 minutes after their event timestamp. Which configuration ensures the late transactions are still incorporated into results without indefinitely delaying window emission?
Setting withAllowedLateness(10 minutes) combined with accumulating-mode triggers lets the pipeline emit an early result at the watermark and then update/re-fire the pane when late data arrives within the allowed lateness window, correctly incorporating late events while still bounding how long the pipeline waits.
A company ingests IoT sensor readings into Bigtable at 50,000 writes/second across 10,000 devices. The current row key format is <timestamp>#<device_id>. Cloud Monitoring shows write throughput is uneven, with some nodes heavily overloaded while others are nearly idle. Which row key redesign BEST resolves this hotspotting?
Placing the high-cardinality device_id first spreads writes evenly across the row-key space and thus across tablets/nodes, since monotonically increasing timestamps as the leading key element concentrate writes on the same tablet range at any given moment.
A retail company's sales data already lives in BigQuery and is updated nightly via scheduled queries. The data science team wants to build a simple linear regression model to forecast next month's revenue per region. They have SQL skills but no experience with Python or ML frameworks, and want to avoid moving data out of BigQuery. Which approach BEST meets these requirements?
BigQuery ML lets the team train and evaluate a linear regression model using only SQL, directly on data already in BigQuery, with no data movement and no need for Python/ML framework expertise.
A company runs a daily Cloud Composer DAG that must complete BigQuery loads by 6 AM for downstream reporting. Recently several tasks silently failed after exhausting retries, and the data team only learned about the missed SLA the next day when analysts complained. Which change BEST ensures the team is notified within minutes of a task failure?
Wiring task failure callbacks into Cloud Monitoring/Pub/Sub and attaching an alerting policy provides near-real-time, proactive notification the moment a task fails, satisfying the SLA-visibility requirement.
A payments company must flag potentially fraudulent transactions within 2 seconds of occurrence. Transactions arrive continuously and unbounded, and the solution must support exactly-once processing with windowed aggregations over sliding time windows. Which architecture BEST satisfies these requirements?
Dataflow with Pub/Sub provides a fully managed, autoscaling streaming architecture with built-in exactly-once semantics and native windowing (fixed, sliding, session), enabling sub-second-to-low-second latency decisioning.
A company runs an on-premises Oracle database and needs near-real-time change data capture (CDC) replication of table changes into BigQuery for analytics, with minimal operational overhead and no custom code to manage.
Datastream is Google Cloud's serverless CDC and replication service, purpose-built to capture change streams from sources like Oracle and MySQL and stream them directly into BigQuery (or GCS) in near real time without managing infrastructure.
A 5-year transaction history table in BigQuery is queried mostly by a transaction_date range, then filtered further by customer_id within that range; occasionally queries filter by customer_id alone across all dates. Which table layout minimizes bytes scanned and cost for these access patterns?
Date-based partitioning prunes irrelevant date ranges for the dominant query pattern, and clustering on customer_id further prunes blocks within the matching partitions, which also helps the occasional customer-only queries by skipping non-matching blocks.
A financial services company serves multiple fraud-detection models in production. They need a centralized way to store, reuse, and serve pre-computed features with low-latency online lookups at prediction time, while also enabling batch feature retrieval for training and monitoring features for drift across teams. Which Google Cloud capability should they use?
Vertex AI Feature Store is purpose-built for this scenario: it provides centralized feature management, low-latency online serving, batch serving for training, feature reuse across teams, and built-in drift monitoring.
A financial services company ingests raw customer support transcripts into Cloud Storage before loading them into BigQuery. Compliance requires that any Social Security numbers or credit card numbers embedded in the free text be identified and redacted prior to loading. Which Google Cloud service should be used to accomplish this requirement?
Cloud DLP (Sensitive Data Protection) provides built-in infoType detectors for SSNs and credit card numbers and can redact or tokenize matched values in unstructured text before it is loaded, directly meeting the requirement.
A 50 TB BigQuery orders table is queried heavily with WHERE clauses filtering on order_date over the last 30 days and equality filters on customer_id. Which table design minimizes bytes scanned and query cost?
Partitioning by order_date lets BigQuery prune partitions outside the 30-day range, and clustering by customer_id further sorts data within each partition so equality filters on customer_id skip irrelevant blocks, minimizing scanned bytes and cost.
A team is migrating nightly batch Spark jobs from on-premises Hadoop to Google Cloud. Each job runs for about 45 minutes once per day. The team wants to minimize cost and avoid any cluster management overhead.
Dataproc Serverless for Spark runs batch Spark workloads without provisioning or managing any cluster; Google Cloud automatically provisions and tears down resources per job, so you pay only for the ~45 minutes of actual compute, matching the cost-minimization and zero-management requirements exactly.
Ready for the real thing?
The full course: two full-length practice tests, video lessons for every exam domain, hands-on labs and detailed explanations.
$129.99 $34.99 with code FREETEST33 — valid through September 16.
Get my $34.99 deal →Professional Data Engineer exam — quick answers
How much does the Professional Data Engineer exam cost?
The exam fee is approximately $200 and varies by region — confirm current pricing with the certification vendor before you book.
What topics are on the exam?
It covers 5 domains: Designing data processing systems (~22%), Ingesting and processing the data (~25%), Storing the data (~20%), Preparing and using data for analysis and ML (~15%), Maintaining and automating data workloads; security and compliance (~18%). The full course has a dedicated chapter, lab and practice-test coverage for each.
Is this practice test really free?
Yes — all questions on this page are free with explanations and no sign-up. The paid Udemy course adds two full-length timed exams, video lessons and hands-on labs.
How do I get the discount?
Use code FREETEST33 at checkout for $34.99 (list $129.99) through September 16 — the enroll button applies it automatically.
Will this prepare me for the real exam?
The questions mirror the real exam's style and are mapped to the official domains. This is exam-focused preparation — combine the free test with the full course's timed simulations to gauge your readiness.