ARA-C01 Question 165
Single answerZones (data warehouse layers)A retail company is redesigning its Snowflake data platform to support batch and near-real-time analytics. The architecture team wants to implement warehouse zones so that raw source data can be retained for replay, business rules can be applied in a controlled layer, and analyst-facing models remain stable even when source systems change. Which design best aligns with Snowflake architectural best practices for data warehouse layering?
- A
Load source data directly into presentation tables and use secure views to preserve raw history when needed.
- B
Create separate RAW, CURATED, and PRESENTATION zones, where RAW stores ingested source-aligned data with minimal transformation, CURATED applies conformance and business rules, and PRESENTATION exposes consumption-ready dimensional or subject-oriented models.
- C
Use a single schema per source system and rely on Time Travel instead of separate zones to recover prior versions of data after transformations.
- D
Store all transformed and raw data in the same tables, but add a STATUS column to indicate whether a record is raw, standardized, or ready for reporting.
Show answer and explanation
Correct answer: B
Explanation
The best answer is the layered zone approach using RAW, CURATED, and PRESENTATION. In Snowflake, these zones are typically implemented through separate databases or schemas depending on organizational standards, but the key principle is logical separation by purpose. RAW data should preserve source-aligned data with minimal changes so teams can audit, replay, and reprocess. CURATED data is where standardization, deduplication, conformance, enrichment, and business rules are applied. PRESENTATION data should provide stable, analytics-ready models for BI tools and data consumers.
This pattern supports core architectural goals tested on the SnowPro Advanced: Architect exam: scalability, maintainability, data lineage, and consumer stability. It also aligns with common Snowflake best practices around separating ingestion from transformation and consumption workloads, reducing blast radius when upstream schemas change, and enabling controlled data governance. While Snowflake features such as Time Travel, cloning, streams, tasks, dynamic tables, and secure views can support implementation, they do not replace the need for clearly defined warehouse layers.
- A. Incorrect.
Incorrect. Loading directly into presentation tables tightly couples source ingestion to consumer-facing models, making downstream analytics fragile when source structures change. Secure views help with controlled access, but they do not replace a raw retention layer or a curated transformation layer. This approach also makes replay, auditability, and incremental reprocessing more difficult.
- B. Correct.
Correct. A layered zone design with RAW, CURATED, and PRESENTATION aligns well with common Snowflake data architecture practices. RAW preserves source fidelity and supports replay or reprocessing. CURATED is the controlled integration layer for cleansing, standardization, conformance, and business logic. PRESENTATION provides stable, consumer-ready structures optimized for analytics, such as dimensional models, marts, or subject-area views. This separation improves maintainability, traceability, and resilience to source-system change.
- C. Incorrect.
Incorrect. Time Travel is useful for recovering changed or deleted data within retention limits, but it is not a substitute for explicit warehouse layering. It does not provide semantic separation between ingestion, business transformation, and consumption layers. A single schema per source system also tends to push transformation concerns into analyst-facing structures and reduces architectural clarity.
- D. Incorrect.
Incorrect. Mixing raw and transformed states in the same physical tables creates governance, lineage, and quality-management problems. A STATUS column does not provide true separation of concerns, and it complicates access control, retention strategy, and downstream consumption. This pattern often leads to inconsistent semantics and harder operational support.