BigQuery Cross-Region Disaster Recovery Strategy
Short answer
D Modify ETL job to load the data into both the current and another backup region to ensure near-real-time replication across regions.
Question
Key Takeaways
Snapshots do not protect against regional failures; synchronous multi-region writes via ETL modification offer the best balance of low RPO and cost efficiency.
Learn how to achieve an RPO of less than 24 hours for BigQuery tables during regional failures while minimizing costs using dual-region ETL loading.
Common mistake: Option A is often chosen because Cloud Storage is cheap, but exporting daily may not sufficiently mitigate the risk of recent data loss compared to continuous replication, and it adds recovery complexity.
Expert Analysis
Regional Failure Protection vs. Regional Snapshots The question specifies a regional dataset and requires protection against regional failures. BigQuery snapshots are point-in-time copies stored within the same region as the source table. As noted in community feedback, "snapshots are stored in the same region, so it dont protect from regional failure." Therefore, Option C fails to meet the core requirement of cross-region resilience.
RPO Constraint Analysis The Recovery Point Objective (RPO) must be less than 24 hours. The data is updated multiple times per day.
- Options A (Export to Cloud Storage) and B (Copy Dataset) are scheduled tasks that typically run once daily. If a failure occurs immediately after the backup runs, you lose up to ~24 hours of data, which technically meets an RPO of <24h only if the window is strictly less than 24 hours, but often implies a full day's gap. More importantly, these are asynchronous backups.
- Option D involves modifying the ETL job to load data into both the current region and another backup region. This creates a synchronous or near-synchronous replication. Since the table is updated multiple times per day, this approach ensures that changes are reflected in both regions almost immediately. This guarantees an RPO significantly lower than 24 hours (potentially minutes or seconds), satisfying the constraint robustly while providing active redundancy.
Cost Considerations While Option A (Cloud Storage) is cheaper for storage, it introduces complexity in recovery (export/import cycles) and may not provide the immediate availability required for business continuity compared to a redundant table. Option B (Dataset Copy) incurs inter-region transfer costs and still relies on a daily schedule. Option D leverages existing infrastructure with minimal overhead beyond the ETL logic change, avoiding additional storage fees for separate backup buckets or datasets while ensuring the highest data fidelity and lowest RPO among the viable options. Community comments highlight that D ensures data is available in both locations nearly in real-time, whereas A and B are limited by their scheduling frequency.
Exam Strategy
When analyzing disaster recovery requirements, always distinguish between intra-region and inter-region protections. For regional failures, intra-region tools like snapshots are insufficient. Evaluate RPO strictness: if 'less than 24 hours' is critical and data updates frequently, synchronous replication (e.g., writing to two regions) is superior to scheduled asynchronous backups (exports/copies) which introduce potential data gaps at the interval boundaries.
Frequently Asked Questions
Why are BigQuery snapshots not suitable for regional disaster recovery?
Is exporting to Cloud Storage cheaper than copying to another region?
How does modifying the ETL job help with RPO?
Community Votes
45% of anonymous learners picked answer A. Votes are pick records left by other test-takers — they are not the verified answer.
Discussion 11 contributions
Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.
From public community threads 11
-
HectorLeon2099 👍 4 Answered AOption A is the most cost efficient: https://cloud.google.com/blog/topics/developers-practitioners/backup-disaster-recovery-strategies-bigquery
-
mdell 👍 2 Answered BIn most cases, it is cheaper to copy a BigQuery dataset to a new region directly rather than exporting it to a Cloud Storage bucket and then loading it into a new BigQuery dataset in the desired region, as you only pay for data transfer costs when copying within BigQuery, while exporting to a bucket incurs additional storage charges for the exported data in Cloud Storage, even if it's only temporary. Key points to consider: No extra storage cost for copying: When copying a BigQuery dataset to a new region, you only pay for the data transfer cost, not the storage of the data in a separate location. Storage cost for exporting: Exporting data to a Cloud Storage bucket means you are charged for the storage of that data in the bucket until you delete it, even if you are just temporarily storing it for transfer.
-
joelcaro 👍 2 Answered DOpción D: Modify ETL job to load the data into both the current and another backup region Evaluación: Ajustar el ETL para escribir en dos tablas (una en la región principal y otra en una región de respaldo) asegura que los datos estén disponibles en ambas ubicaciones casi en tiempo real. Esto garantiza un RPO de menos de 24 horas, ya que las actualizaciones intradía se reflejan en ambas regiones. Aunque podría aumentar los costos de almacenamiento por duplicar los datos, es la solución más efectiva y directa para proteger contra fallos regionales.
-
gabbferreira 👍 2snapshots are stored in the same region, so it dont protect from regional failure
-
AwesomeGCP 👍 1 Answered BThe correct answer is B. Schedule a daily copy of the dataset to a backup region. Why this is the best choice: Protects Against Regional Failures & Meets RPO: Copying the dataset to another region once daily creates a standby copy in a completely independent failure domain, satisfying the cross-region disaster recovery requirement within an RPO of less than 24 hours. Keeps Costs to a Minimum: Cross-region dataset replication and BigQuery dataset copies are incremental. Running a daily scheduled copy transfers and stores only the necessary data blocks, avoiding active duplicate compute/ETL pipelines and external storage management costs. Native Query Readiness: Because the target is a BigQuery dataset in the secondary region, failover can occur immediately with zero ingestion or reload overhead if the primary region becomes unavailable. Why the other options are incorrect: A (Daily export to a Cloud Storage dual/multi-region bucket): Exporting to GCS introduces additional complexity and recovery time objective (RTO) overhead. If the primary BigQuery region fails, you must ingest or load the files back into a BigQuery dataset in another region before querying, and you pay both export compute and persistent GCS storage costs. C (Daily BigQuery snapshot): BigQuery table snapshots are strictly bound to the same region as the base table. They protect against accidental table corruption or deletion, but provide zero resilience against an entire regional failure. D (Dual-region ETL ingestion): Modifying your pipeline to ingest and transform data into two regions concurrently doubles pipeline execution resources, ingestion bandwidth, and compute costs continuously throughout the day—violating the requirement to keep costs to a minimum.
Show 6 more comments
-
MarcoPellegrino 👍 1 Answered Dhttps://cloud.google.com/blog/topics/developers-practitioners/backup-disaster-recovery-strategies-bigquery Google presents both A and D Why A: - Cost: Lower. GCS storage is significantly cheaper than BigQuery storage. You pay for storage in GCS and minimal egress charges when exporting. - Complexity: Simpler. You schedule a daily export job. Restoring involves importing from GCS to BigQuery in another region. - Consistency: Easier to manage. The export process creates a consistent snapshot of the data at the time of export. You might have some latency (up to 24 hours in this scenario), but the data within the export is consistent. - RPO: Meets the requirement. A daily export ensures an RPO of less than 24 hours. - RTO: Depends on the restore process from GCS to BigQuery. You can pre-provision slots in the backup region to minimize restore time.
-
Parandhaman_Margan 👍 1 Answered CMeets the RPO requirement (< 24 hours) Cost-effective solution Quick recovery from regional failures
-
mdell 👍 1Additionally it only mentions backing up the sales table and not the entire dataset
-
desertlotus1211 👍 1 Answered Aalmost the same as 211, 211 says multi-region vs regional...
-
FireAtMe 👍 1 Answered ABoth A and B works. But it is cheaper to save data in GCS.
-
22c1725 👍 1 Answered ANot 'C' snapchot are stored in the same region.
Reader comments & corrections
No reader corrections yet — spotted an error or have a note? Share it below.