Snowflake Overview & Key Concepts

Overview

Snowflake is a fully managed, cloud-native data platform provided as Software-as-a-Service (SaaS). For the SnowPro Core (COF-C03) exam, understanding Snowflake's fundamental architecture and value proposition is critical, as it forms the foundation for all other topics in Domain 1 (Snowflake AI Data Cloud Features & Architecture), which represents 31% of the exam.

Snowflake was built from the ground up for the cloud, differentiating itself from traditional on-premises data warehouses that were simply "lifted and shifted" to cloud infrastructure.

Key Concepts

Software as a Service (SaaS)

Snowflake is a true SaaS offering. This means:

• No Hardware to Install: There is no physical hardware to buy, configure, or manage.
• No Software to Install: There is no software to install, update, or patch. Snowflake handles all maintenance and upgrades transparently.
• Fully Managed: Snowflake manages the infrastructure, performance tuning, data protection, and availability.

Cloud-Agnostic

Snowflake can be deployed on any of the three major public clouds:

• Amazon Web Services (AWS)
• Microsoft Azure
• Google Cloud Platform (GCP)

This allows organizations to avoid vendor lock-in and adopt a multi-cloud strategy while maintaining a consistent data platform experience.

Multi-Cluster Shared Data Architecture

Snowflake's architecture is unique because it combines the best of two traditional architectures:

• Shared-Disk Architecture: Multiple nodes access the same centralized storage. Advantage: Single source of truth. Disadvantage: Contention when scaling compute.
• Shared-Nothing Architecture: Each node has its own local CPU and storage. Advantage: Scales out well. Disadvantage: Requires data redistribution (shuffling) when scaling.

Snowflake uses a hybrid approach: Multi-cluster shared data architecture. It has a central persistent storage layer (like shared-disk) that is accessed by multiple independent compute clusters (like shared-nothing).

How It Works

Snowflake's architecture is divided into three logically integrated but physically separated layers:

1. Storage Layer (Database Storage): A centralized repository for all data, stored in cloud object storage (e.g., Amazon S3, Azure Blob, GCS).

2. Compute Layer (Query Processing): Virtual Warehouses (MPP compute clusters) that execute queries against the data in the storage layer.

3. Cloud Services Layer: The "brain" that coordinates everything (authentication, metadata, query optimization, caching).

Separation of Storage and Compute

The most important architectural feature of Snowflake is the complete separation of compute and storage.

• They scale independently. You can add more compute without adding storage, and vice-versa.
• They are billed independently.
• Multiple compute clusters can access the same centralized storage simultaneously without contention.

SQL Examples

-- Check your current Snowflake version
SELECT CURRENT_VERSION();

-- Check the cloud region your account is hosted in
SELECT CURRENT_REGION();

-- View your current account name
SELECT CURRENT_ACCOUNT();

Exam Tips

⚠️ Crucial for COF-C03:

• Remember that Snowflake does not run on on-premises infrastructure. It is cloud-only.
• Understand the difference between Shared-Disk and Shared-Nothing, and how Snowflake is a hybrid.
• The phrase "Separation of Storage and Compute" is the answer to many architecture benefit questions (e.g., "Why can multiple teams run heavy queries without impacting each other?").
• Snowflake handles all maintenance, indexing (micro-partitioning), and tuning.

Key Takeaways

• Snowflake is a true SaaS data platform.
• It operates on AWS, Azure, and GCP.
• Its core innovation is the Multi-cluster Shared Data Architecture (hybrid of shared-disk and shared-nothing).
• Storage and compute are completely decoupled, scaling and billing independently.