Storing the data
Free Google Cloud Certified - Professional Data Engineer practice — 6 questions on Storing the data, with explanations. No sign-up. Full 12-question mixed test →
A company ingests IoT sensor readings into Bigtable using a row key composed solely of a Unix timestamp (e.g., '1706745600'). During peak ingestion, most writes land on a single tablet server, causing severe hotspotting even though the cluster has many nodes. Which row key redesign BEST resolves this?
Hashing (or otherwise distributing) a high-cardinality field like sensor_id and placing it before the timestamp spreads writes across the key space and thus across tablets, eliminating the hot tablet caused by monotonically increasing timestamp keys. This is the standard Bigtable anti-hotspotting pattern tested on the exam.
A 50TB BigQuery fact table 'orders' grows daily. Nearly every analytical query filters on order_date over a rolling 90-day range and filters on customer_region (medium cardinality, ~50 values); a smaller subset of queries also filters on product_category (high cardinality) within that region/date scope. Which table design BEST optimizes cost and performance for this access pattern?
Partitioning by order_date lets BigQuery prune partitions outside the 90-day window before scanning, which every query benefits from. Clustering by customer_region first (used by nearly all queries) and then product_category (used by a subset) further sorts blocks so both filters can prune data, maximizing bytes-scanned savings.
A media company stores web server logs in Cloud Storage. Access patterns are unpredictable: some log files are queried heavily for weeks, others go cold within days. Logs must be retained for 7 years for compliance, and legal requires that retained objects cannot be deleted or overwritten early, even by an administrator. Which configuration BEST meets these requirements while minimizing storage cost?
Autoclass automatically moves individual objects between Standard, Nearline, Coldline, and Archive based on actual access patterns rather than a fixed age, which is ideal when access is unpredictable — it saves cost without manual tuning. Bucket Lock enforces an immutable retention policy so objects genuinely cannot be deleted or overwritten before the 7-year period, satisfying the legal requirement — something plain lifecycle rules or storage class choice cannot do.
You are designing the backend for a global multiplayer game that must maintain strongly consistent, externally consistent player-state transactions across multiple continents, support complex relational joins, and scale horizontally to petabytes with a 99.999% availability SLA. Which storage service should you choose?
Cloud Spanner is Google's globally distributed, horizontally scalable relational database offering external consistency (true serializability) across regions, SQL joins, and up to 99.999% availability with a multi-region configuration — exactly matching all stated requirements.
A dataset's time travel window is configured at the default 7 days. A critical production table was accidentally dropped 10 days ago and nobody noticed until now. What is the correct statement about recovering this table's data?
After the configurable time travel window (default 7 days) expires, BigQuery enters an additional fail-safe period (up to 7 more days) during which deleted/changed data is not queryable via SQL but can still potentially be recovered by engaging Google Cloud Support — this is a specific, exam-tested nuance beyond plain time travel.
A retail analytics team runs BigQuery workloads that are nearly idle most months but spike to require massive concurrent slot capacity for 3-4 days during quarterly batch reprocessing. Finance wants to avoid a long-term annual commitment while still guaranteeing capacity during the spikes. Which approach BEST fits these requirements?
Flex-rate / short-term slot commitments (available through BigQuery editions and reservations) can be purchased for as little as 60 seconds up to a defined duration, letting you guarantee slot capacity exactly during the multi-day spike and cancel afterward — avoiding both unpredictable on-demand cost spikes and long-term (annual/monthly) commitment waste. Autoscaling reservations further let capacity flex within that window.
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 →