ARA-C01 Question 182
Single answerDeploymentA global retail company is deploying Snowflake across AWS, Azure, and GCP to support regional data residency requirements and low-latency analytics for local business units. The architecture team wants a deployment approach that minimizes cross-region data movement, allows each region to continue operating if another region is unavailable, and provides a practical way to share a subset of centrally governed product reference data to all regions. Which approach best meets these requirements?
- A
Deploy a single Snowflake account in one cloud region and use reader accounts for all other regions so local teams can query the central data without replication.
- B
Deploy separate Snowflake accounts in each required region/cloud, use database replication and failover groups where supported for business continuity, and distribute shared reference data using replication or data sharing patterns appropriate to the regional topology.
- C
Deploy one organization account and place all regional workloads in separate virtual warehouses, because virtual warehouses provide isolation and regional independence without requiring additional accounts.
- D
Deploy separate databases for each region inside one Snowflake account and rely on Time Travel and Fail-safe to maintain availability if an entire cloud region becomes unavailable.
Show answer and explanation
Correct answer: B
Explanation
The key deployment requirements are regional data residency, low-latency local analytics, reduced cross-region movement, and resilience if another region is unavailable. In Snowflake, these goals are best addressed by a distributed account strategy: deploy separate Snowflake accounts in the necessary regions and clouds, keep local workloads local, and use Snowflake replication capabilities for selected databases and supported account-level objects. Where business continuity is required, failover groups are the primary mechanism for replicating and failing over supported objects between accounts in different regions or clouds, subject to Snowflake edition and feature support. For sharing centrally governed reference data, architects commonly use secure data sharing when topology allows, or replication to place read-mostly shared data closer to consuming regions. By contrast, a single account with warehouses or databases does not create geographic independence, and recovery features like Time Travel and Fail-safe do not replace cross-region deployment or disaster recovery design. Relevant Snowflake documentation includes sections on account architecture, database replication, replication groups/failover groups, business continuity and disaster recovery, organizations, and data sharing.
- A. Incorrect.
Incorrect. Reader accounts are primarily intended to let third parties consume shared data without becoming full Snowflake customers. They do not solve multi-region or multi-cloud deployment requirements for independent regional operations, and they do not provide regional isolation or local processing in another cloud/region. A single centrally hosted account would still concentrate compute and storage in one region, increasing latency and violating the goal of minimizing cross-region data movement.
- B. Correct.
Correct. For multi-region and multi-cloud deployment with data residency and operational independence, the practical architecture is to deploy separate Snowflake accounts in the required regions/clouds. Snowflake supports replication of databases and account objects, and failover groups can be used to replicate and fail over supported objects for business continuity across regions/cloud platforms depending on edition and topology. Centrally governed reference data can then be distributed to regional accounts using replication and/or secure data sharing patterns that fit the deployment design. This approach aligns with Snowflake best practices for geographic distribution, resilience, and localized workload execution.
- C. Incorrect.
Incorrect. An organization in Snowflake is a management boundary for accounts, billing, and governance features; it is not a compute deployment construct that replaces regional accounts. Virtual warehouses isolate compute workloads within an account, but they do not create separate regional deployments, do not satisfy cross-cloud or cross-region residency needs, and do not provide survivability if the underlying region hosting the account is unavailable.
- D. Incorrect.
Incorrect. Separate databases within one account can help logically isolate data, but they remain dependent on the same account and its underlying cloud region. Time Travel and Fail-safe are data recovery features, not regional disaster recovery mechanisms. They do not keep workloads running if an entire region is unavailable, nor do they reduce latency for globally distributed users.