ARA-C01 Question 291
Single answerArchitecture changesA global retail company currently runs all Snowflake workloads in a single account and a single AWS region. Over the last year, the platform has expanded to support independent data engineering, BI, and data science teams. The company is experiencing the following issues: production ETL jobs occasionally consume resources needed by executive dashboards, data science users require broader experimentation without risking production objects, and a newly adopted CI/CD process needs a safer way to promote database changes across environments. Leadership wants an architectural change that improves workload isolation and environment governance while minimizing unnecessary operational complexity. Which approach best addresses these requirements?
- A
Keep a single Snowflake account, create separate virtual warehouses for ETL, BI, and data science, and use role-based access control to distinguish development and production work.
- B
Adopt a multi-account strategy with separate Snowflake accounts for development/test and production, and use separate virtual warehouses within each account for ETL, BI, and data science workloads.
- C
Move all workloads into one larger virtual warehouse with multi-cluster enabled so that ETL, BI, and data science can share compute elastically across environments.
- D
Create separate databases for development, test, and production in the same account, and rely on zero-copy cloning instead of changing the account architecture.
Show answer and explanation
Correct answer: B
Explanation
The scenario points to two distinct architectural pressures: compute isolation and environment isolation. In Snowflake, separate virtual warehouses are the primary mechanism for isolating compute across workloads such as ETL, BI, and data science. This prevents one workload class from consuming compute intended for another and supports more predictable performance and cost management. At the same time, when organizations mature and implement CI/CD, separate Snowflake accounts for development/test and production are commonly adopted to strengthen governance, reduce accidental impact to production, and create clearer lifecycle boundaries.
A single-account architecture can be appropriate early on, especially when using separate warehouses and strong RBAC. However, as requirements evolve, architects often introduce account-level separation for regulated, high-risk, or operationally mature environments. Multi-account strategies are also consistent with Snowflake guidance around organizational isolation and environment separation, while still allowing centralized management through Snowflake Organizations, account replication/failover where applicable, and deployment automation.
Option 2 is therefore the best architectural change because it addresses all stated problems without overcomplicating the design: separate accounts for dev/test versus production, plus separate warehouses for ETL, BI, and data science within each account. Relevant Snowflake best practices include workload isolation with virtual warehouses, least-privilege access design with RBAC, and using separate environments/accounts when stronger governance and operational separation are required.
- A. Incorrect.
This option partially addresses compute contention by separating ETL, BI, and data science onto different virtual warehouses, which is a Snowflake best practice for workload isolation. However, it does not sufficiently address environment governance and safer promotion of database changes across environments. Keeping development/test and production in the same account increases the risk of administrative mistakes, accidental access to production objects, and weaker isolation for CI/CD processes. RBAC helps, but it is not a full substitute for stronger account-level separation when teams and environments have expanded significantly.
- B. Correct.
This is the best answer because it combines two architectural improvements that directly map to the scenario. First, using separate accounts for development/test and production provides stronger environment isolation, which supports safer CI/CD promotion patterns and reduces risk to production. Second, using separate virtual warehouses within each account isolates ETL, BI, and data science compute so one workload does not starve another. This design aligns with Snowflake architectural best practices: use separate warehouses for independent workloads and consider multiple accounts when stronger isolation, governance, or lifecycle separation is needed.
- C. Incorrect.
This option is incorrect because a larger multi-cluster warehouse mainly helps concurrent queries on the same workload pattern, especially BI-style concurrency, but it does not provide proper separation between different workload types or environments. ETL, BI, and data science have different performance and cost characteristics, and sharing one warehouse can still make usage governance and chargeback more difficult. It also does not solve the need for safer environment promotion across development/test and production.
- D. Incorrect.
This option is a plausible intermediate design, but it is not the best fit for the stated requirements. Separate databases for development, test, and production in one account can help organize objects, and zero-copy cloning is useful for creating lower-environment copies. However, this still lacks strong account-level isolation for governance and CI/CD safety. It also does not fully address the broader architectural need created by team growth and production risk. Database separation is weaker than account separation when the goal includes reducing blast radius and improving environment governance.