PDE domain - 20% of the exam

Storing Data

Storing Data is 20% of the PDE exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

64
Practice questions
4
Objectives
30
Medium
34
Hard

Where people go wrong

  • Reversing the roles of Bigtable and BigQuery

    The wrong answers swap the two services, give Bigtable BigQuery's columnar analytical layout, or treat slot-based execution as a Bigtable feature. Bigtable is a wide-column store sorted by row key, so a key with the device first and time second turns a per-device lookup of recent readings into a small local read at low latency. BigQuery is the columnar warehouse that scans very large tables for aggregates, and it is not the home for keyed operational lookups.

    See it in a sample question
  • Treating regional databases as multi-region writers

    The wrong answers offer AlloyDB, Cloud SQL or a larger instance for a ledger or inventory that needs strongly consistent writes across regions. AlloyDB has a regional primary with asynchronous cross-region read pools, and Cloud SQL has per-instance limits with asynchronous replicas, so neither gives read-after-write across regions without manual sharding. Cloud Spanner in a multi-region configuration provides synchronous replication and transactions across rows in different regions, which removes the sharding and compensating logic.

    See it in a sample question
  • Putting binary media in a database

    The wrong answers place video or photos in Firestore, Bigtable or Cloud SQL. Firestore documents are capped at one mebibyte, Bigtable is tuned for small cells and high-throughput rows and is not an object store, and Cloud SQL is a poor blob store at scale. Cloud Storage holds the bytes, works with HTTP caching and CDNs, and offers lifecycle rules that move objects to Nearline and Archive as access cools. Firestore keeps only the profile document, with a reference to the object path.

    See it in a sample question
  • Mistaking external tables or copies for BigLake

    The wrong answers keep standard external tables, or treat BigLake as a billing change or a hidden copy. Standard external tables lack row-level and column-level security and need bucket read access for each querying principal. A BigLake table reads the files in place through a connection whose service account holds the bucket permission, so querying users need only BigQuery access, BigQuery enforces the row and column policies, and connected engines such as Spark see the same controls.

    See it in a sample question
  • Sharing across organisations with views or copies

    The wrong answers share data with authorised views, direct IAM grants, scheduled copies or exported files. These run compute in the publisher's project, duplicate storage, or give no per-subscriber entitlement. Analytics Hub publishes listings in a data exchange, each subscriber creates a read-only linked dataset in their own project, queries run on the subscriber's slots, and the publisher revokes one subscription without affecting the others.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleStoring Datahard

An architect is comparing BigQuery and Bigtable for a workload that records device telemetry from two million industrial sensors. Each sensor emits a reading every second, and downstream applications need to retrieve the most recent 24 hours of readings for any single sensor within tens of milliseconds, while a separate weekly analytical job scans aggregates across the full fleet. The architect wants to understand the fundamental role boundary between the two services. Which statement most accurately describes how BigQuery and Bigtable differ for this workload?

  • ABigtable is a wide-column NoSQL store with sorted row keys that gives single-digit millisecond reads for a known key, while BigQuery is a columnar analytical warehouse designed for high-throughput scans across very large tables; the per-sensor lookup belongs in Bigtable and the weekly aggregate belongs in BigQuery. Correct
  • BBigQuery is a wide-column NoSQL store optimised for single-row lookups by key, while Bigtable is a columnar analytical warehouse tuned for ad hoc SQL scans, so the per-sensor lookups should target BigQuery and the weekly aggregate should target Bigtable.
  • CBigQuery and Bigtable both target operational workloads, but BigQuery is preferred whenever rows exceed one kilobyte and Bigtable is preferred whenever rows are smaller, regardless of access pattern.
  • DBigtable and BigQuery are interchangeable for telemetry because both are columnar; the team should pick the cheaper one for the region and accept identical latency characteristics from each service.
Distinguish Bigtable as a low-latency wide-column NoSQL store from BigQuery as a columnar analytical warehouse when serving telemetry. Bigtable stores rows sorted by a single row key and is engineered for low-latency point and small range reads at very high write rates, which is exactly the per-sensor recent-history pattern. BigQuery stores data in columnar format across distributed storage and uses a slot-based execution engine that excels at scanning and aggregating across large tables, which is the weekly fleet-wide pattern. Choosing each service for the access pattern it was built for is the canonical PDE role boundary.

Why A is correct: Bigtable is sorted by row key and serves point and small range reads in low single-digit milliseconds, which suits the per-sensor 24-hour lookup, while BigQuery's columnar storage and slot-based execution are designed to scan and aggregate across very large tables on schedule, which suits the weekly cross-fleet job.

Why B is wrong: This reverses the actual roles. BigQuery is the columnar analytical warehouse and Bigtable is the wide-column key-ordered NoSQL store, so the description swaps the two services. A candidate who only half-remembers the column orientation of BigQuery can fall into this trap.

Why C is wrong: BigQuery is an analytical warehouse, not an operational store, and the selection between Bigtable and BigQuery is driven by access pattern rather than row size. The size-based rule sounds concrete but is fabricated and will mislead a candidate who has not internalised the role boundary.

Why D is wrong: Although both services use a column-oriented physical layout, their access patterns and latency profiles are very different. Bigtable serves low-latency keyed reads while BigQuery serves throughput-oriented scans, so they are not interchangeable for a real-time per-sensor lookup.

Free sampleStoring Datahard

A global ride-hailing company is designing a new driver-payments ledger that must serve riders and drivers in 14 regions across four continents. The schema centres on a payments table with an expected sustained write rate of around 60,000 transactions per second at launch, projected to triple within 18 months. Every payment requires a strongly consistent debit and credit across two accounts that may live in different regions, with read-after-write guarantees for both rider and driver apps. The team wants to avoid building manual sharding or compensating-transaction logic. Which storage system best meets these requirements?

  • ADeploy AlloyDB for PostgreSQL in a primary region with cross-region read pools and route writes through the primary using two-phase commit across logical accounts.
  • BProvision Cloud SQL for PostgreSQL with high availability in the busiest region and use logical replication to read replicas in the other regions for nearby reads.
  • CUse Bigtable with a row key that concatenates account_id and timestamp, applying single-row mutations for each leg of the payment to achieve atomicity.
  • DUse Cloud Spanner in a multi-region configuration and model payments as a single transaction across the rider and driver account rows. Correct
Recognise that strongly consistent multi-region transactional workloads at high throughput belong on Spanner rather than Cloud SQL, AlloyDB, or Bigtable. Spanner is the only Google Cloud database that combines horizontally scalable writes, synchronous multi-region replication, and external consistency through TrueTime, which together allow a single transaction to atomically update rows that live in different regions while preserving read-after-write semantics for any client.

Why A is wrong: AlloyDB is a regional service and its cross-region read pools are asynchronous, so writes still funnel through a single region and cross-region transactions would require application-level two-phase commit, exactly the logic the team wants to avoid.

Why B is wrong: Cloud SQL caps out at a single regional primary and cannot deliver strongly consistent multi-region writes at tens of thousands of TPS, so it cannot meet the launch throughput or the read-after-write guarantee for users in distant regions.

Why C is wrong: Bigtable only offers atomicity at the single-row level, so a debit and credit on two different account rows cannot be made atomic, which breaks the ledger's correctness requirement even though throughput would be plentiful.

Why D is correct: Spanner provides horizontally scalable, externally consistent transactions across regions with synchronous replication, so a single read-write transaction can debit and credit accounts in different regions and offer read-after-write guarantees without manual sharding.

Free sampleStoring Datahard

A media company ingests roughly 40 TB of raw video files per month from camera crews and processes them through a transcoding pipeline that reads the originals and writes multiple renditions. Originals are accessed heavily for two weeks during editing, then mostly archived but occasionally needed for re-edits up to three years later. Access patterns are object reads and writes by file name, no transactional updates are required, and storage cost over the lifecycle dominates the bill. Which design should the team adopt?

  • AStore originals in Cloud Storage Standard with an Object Lifecycle Management rule that transitions objects to Nearline after 14 days and to Archive after 90 days. Correct
  • BStore originals in Firestore as binary fields under a per-project document, relying on Firestore document TTL for archival.
  • CStore originals in Bigtable with the file name as row key and the bytes split across column families, using garbage collection to expire old versions.
  • DStore originals in Cloud SQL with a BYTEA column per file and partition tables monthly, dropping old partitions after three years.
Match large opaque blob workloads with predictable lifecycle access to Cloud Storage with class transitions rather than to document, columnar, or relational stores. Cloud Storage is the standard blob store on Google Cloud and is the only listed option that scales economically for terabytes of large opaque files. Object Lifecycle Management automates transitions from Standard to Nearline, Coldline, or Archive based on object age, which lines up directly with the described hot-then-cold access pattern and minimises total cost without sacrificing durability or retrieval ability.

Why A is correct: Cloud Storage is the appropriate blob store for large video files, and lifecycle rules move objects through storage classes as access cools, matching the two-week hot window and the multi-year cold tail at the lowest total cost.

Why B is wrong: Firestore is a document database whose documents are capped at one mebibyte and is not designed for multi-gigabyte binary payloads, so storing video originals there is technically infeasible and economically nonsensical.

Why C is wrong: Bigtable is a wide-column key-value store tuned for high-throughput small reads and writes, not multi-gigabyte blobs, and it would inflate cost and complexity while providing no benefit over an object store for this access pattern.

Why D is wrong: Cloud SQL can technically hold binary data, but relational engines are poor blob stores at this scale, vertical sizing becomes prohibitive at tens of terabytes per month, and there is no lifecycle tiering analogous to Cloud Storage classes.

Free sampleStoring Datahard

A team has Avro and Parquet files in Cloud Storage that are currently queried through BigQuery external tables. They want to enforce column-level security, integrate with the BigQuery metastore, and allow open-source query engines such as Spark to read the same data with the same governance model. Which option best describes how BigLake changes their setup compared with the existing external tables?

  • ABigLake tables are identical to external tables in capability and only change the billing model, so the team should expect no governance improvements when migrating from external to BigLake tables.
  • BBigLake tables copy the underlying files into a managed Cloud Storage bucket owned by BigQuery, after which row-level and column-level security become available because the files are no longer external.
  • CBigLake tables disable access from open-source engines and route all queries through BigQuery, so the team would have to retire their Spark jobs once they migrate.
  • DBigLake tables replace external tables with a managed connection that delegates storage access through a service account, allowing BigQuery to enforce row-level and column-level security and letting compatible open-source engines read the same data under the same policies. Correct
Recognise BigLake as the governance layer that extends BigQuery security to open-format files while keeping them readable by other engines. A BigLake table is a BigQuery table type backed by files in Cloud Storage and a connection resource that holds the identity used to read those files. Because access flows through that connection rather than through the user's own permissions on the bucket, BigQuery can apply row-level and column-level security and column masking on top of the open-format files. The same BigLake tables can be read by Spark and other engines through connectors that respect the BigLake policy enforcement, which is the unified governance story the team is looking for.

Why A is wrong: External tables in BigQuery do not support row-level or column-level security in the way BigLake tables do, and BigLake adds a connection-based access model that external tables lack. Treating BigLake as a billing-only change misses its central governance value.

Why B is wrong: BigLake does not copy the files; the data continues to live in the customer's Cloud Storage bucket and is read through a connection. Coupling fine-grained security to a hidden copy step is a believable but incorrect mental model.

Why C is wrong: A defining purpose of BigLake is to keep the data accessible to open-source engines such as Spark through connectors while still applying the BigLake governance layer. Claiming that BigLake locks out non-BigQuery engines contradicts the published architecture.

Why D is correct: BigLake tables introduce a connection resource that holds a service account used to read the files, so BigQuery can enforce fine-grained security on top of object storage. The same BigLake metadata is exposed to open-source engines through connectors, giving consistent governance across query engines.

Free sampleStoring Datahard

A financial services group wants its risk team to publish a curated set of BigQuery views to several external partner organisations and to two internal subsidiaries that operate in different Google Cloud organisations. The risk team must keep the underlying tables in their own project, must not duplicate storage, must control which subscribers see which views, and must be able to revoke access without rewriting downstream queries. Which Google Cloud service is designed for this cross-organisation data sharing pattern?

  • ABigQuery authorised views shared directly with each partner's service account, which lets subscribers query the views without exposing the underlying tables and which can be revoked through IAM.
  • BBigQuery Data Transfer Service jobs that copy the curated views into each subscriber's project on a schedule, so each subscriber owns a fresh local copy of the data each day.
  • CCloud Storage signed URLs over an export of the BigQuery views, distributed to each subscriber on a recurring schedule so they can load the data into their own warehouses.
  • DAnalytics Hub, where the risk team publishes a data exchange with listings backed by their BigQuery views, and each subscriber creates a linked dataset in their own project without copying data. Correct
Recognise Analytics Hub as the BigQuery service that exchanges curated datasets across organisations without copying data. Analytics Hub provides a publisher and subscriber model on top of BigQuery. A publisher creates a data exchange and listings backed by datasets or views, and each subscriber creates a linked dataset in their own project. Linked datasets are read-only pointers that incur no storage duplication, queries continue to work after revocation only while entitlements hold, and the publisher manages entitlements centrally, which fits the cross-organisation no-copy requirement.

Why A is wrong: Tempting because authorised views do hide base tables and use IAM, but managing per-subscriber service accounts across many organisations does not scale and lacks the listing, subscription, and entitlement model that the scenario requires.

Why B is wrong: Tempting because Data Transfer Service does move BigQuery data on a schedule, but it duplicates storage in every subscriber project, which directly violates the no-duplication requirement in the scenario.

Why C is wrong: Tempting as a simple cross-organisation distribution channel, but it requires exports, duplicates storage at every subscriber, breaks downstream SQL when files change, and lacks central entitlement control.

Why D is correct: Correct because Analytics Hub is the BigQuery-native exchange that publishes listings, lets subscribers create read-only linked datasets in their own projects with no storage duplication, and lets the publisher revoke entitlements centrally.

Other domains in this exam

See also the PDE cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Google Cloud. Original, blueprint-aligned practice material only.