Answer: Micro-partitions, Persisted query results

The correct answers are **B: Micro-partitions** and **D: Persisted query results**. ### Explanation of Snowflake's Architecture Layers Snowflake uses a **three-layer architecture** that separates storage, compute, and cloud services for scalability and performance: - **Storage Layer** (persistent data in cloud object storage like S3/Azure Blob/GCS): This layer holds all durable, user data in a compressed, columnar, immutable format. It is independent of compute and scales automatically. - **Compute Layer** (virtual warehouses): Handles query execution, using local SSD caches for performance. It pulls data from the storage layer as needed. - **Cloud Services Layer**: Manages metadata, query optimization, authentication, access control, and various caches (including the **query result cache**). ### Why B and D? - **Micro-partitions (B)**: These are the fundamental units of storage in Snowflake. All table data is automatically divided into micro-partitions (typically 50–500 MB uncompressed, stored compressed). They enable efficient pruning (skipping irrelevant data) and columnar access. This is core persistent data stored directly in the **storage layer**. - **Persisted query results (D)**: Snowflake maintains a **result cache** (also called persisted query results) for up to 24 hours (or longer in some contexts with certain conditions). When a query is re-run with no changes to the underlying data or session parameters, Snowflake can serve results directly from this cache without re-executing the query or scanning micro-partitions. While closely tied to the services/compute layers for access, the persisted results themselves are treated as stored data in exam contexts under the broader storage considerations. ### Why not the others? - **A: Snowflake parameters** — These are configuration settings (e.g., session or account parameters) managed in the **cloud services layer** (metadata), not the storage layer. - **C: Query history** — Stored and managed in the **cloud services layer** (for auditing, monitoring, and INFORMATION_SCHEMA views). It is metadata, not persistent table data. - **E: Standard and secure view results** — Views are logical definitions (metadata in cloud services). Their results are computed on-the-fly or pulled from underlying table micro-partitions in the storage layer. There is no separate "view results" storage. This question tests your understanding of where different elements live in Snowflake's decoupled architecture. The storage layer focuses on durable table data (micro-partitions) and associated persisted elements like query results for efficiency. **Tip for the exam**: Remember that user data lives in micro-partitions in storage, while most "history," "parameters," and "metadata" live in cloud services. Caching (results, local warehouse cache) optimizes performance but ties back to storage for the base data.