ARA-C01 Question 166
Single answerZones (data warehouse layers)A retail company is redesigning its Snowflake data warehouse to support governed self-service analytics and repeatable data pipelines. The architecture team wants to organize data into zones so that raw source data is preserved, transformations are traceable, and business users consume curated datasets with stable definitions. Source files arrive from multiple operational systems with inconsistent schemas and occasional data quality issues. Which approach BEST aligns with Snowflake architectural best practices for implementing data warehouse zones?
- A
Load source files directly into presentation tables used by BI tools, and use views on top of those tables only when schema mismatches are discovered.
- B
Create a raw landing zone for immutable source data, a standardized/integration zone for cleansing and conformance, and a presentation zone for curated business-ready models consumed by downstream users.
- C
Store all data in a single normalized schema and use separate virtual warehouses to represent raw, curated, and presentation zones.
- D
Use transient tables for all zones so that Time Travel and Fail-safe costs are minimized, then rely on BI tools to apply business rules at query time.
Show answer and explanation
Correct answer: B
Explanation
In Snowflake, zones or layers are a logical architectural pattern rather than a special built-in feature. A well-designed warehouse commonly separates data into at least: (1) a raw/landing zone that preserves ingested source data with minimal transformation, (2) an integration/standardized zone where cleansing, conformance, deduplication, and business rules are applied, and (3) a presentation/consumption zone that exposes trusted, business-ready datasets for analytics. This approach improves governance, lineage, reproducibility, and change isolation. It also aligns with Snowflake best practices of separating storage and compute concerns: virtual warehouses should be used for workload isolation and performance management, while schemas/databases/tables/views implement the data layers. Snowflake documentation and architecture guidance consistently emphasize preserving raw data for traceability, using ELT patterns for transformation within Snowflake, and presenting curated data products to end users rather than exposing ingestion structures directly.
- A. Incorrect.
This is not the best practice for a zoned warehouse architecture. Loading directly into presentation tables couples ingestion with consumption, makes lineage and reprocessing harder, and exposes business users to source-system volatility and data quality issues. A presentation layer should generally contain curated, stable datasets rather than raw operational loads.
- B. Correct.
This is the best answer. A common Snowflake architectural pattern is to separate data into logical layers or zones: a raw/landing zone to preserve source fidelity, an integration or standardized zone to cleanse, conform, and apply transformation logic, and a presentation/consumption zone to expose business-ready data products. This design supports auditability, replay/reprocessing, controlled data quality, and stable semantic consumption.
- C. Incorrect.
This reflects a misconception about zones in Snowflake. Virtual warehouses provide compute isolation, not logical data layering. Zones are implemented through database/schema/table/view design and pipeline flow, not by assigning different compute clusters. A single normalized schema also does not address the need to preserve raw data and publish curated models separately.
- D. Incorrect.
This is incorrect because making every zone transient is usually inappropriate. Raw and curated layers often require stronger recovery, lineage, and retention characteristics, and transient objects reduce data protection by removing Fail-safe. Also, pushing business rules into BI tools undermines centralized governance, creates inconsistent metrics, and weakens the value of a curated presentation zone.