Understanding Snowflake Architecture: Storage, Compute, and Cloud Services Explained

Learning Snowflake Series | Topic 3

In my previous articles, I explored why Snowflake has become so popular and how it differs from traditional data warehouses.

While learning Snowflake, I kept hearing one statement again and again:

“Snowflake separates storage and compute.”

At first, I understood the concept at a high level.

But I wanted to know what was actually happening behind the scenes.

That’s when I started exploring Snowflake’s architecture.

The more I learned about it, the more I realized that understanding the architecture makes many other Snowflake concepts much easier to understand.

Let’s break it down in a simple way.

What is Snowflake Architecture?

Snowflake follows a unique architecture that separates data storage, data processing, and system management into different layers.

Instead of putting everything into a single system, Snowflake divides responsibilities across three layers.

This design helps improve scalability, performance, and flexibility.

The three layers are:

1. Storage Layer

2. Compute Layer

3. Cloud Services Layer

Each layer has a specific responsibility, and together they form the foundation of Snowflake’s Data Cloud.

Storage Layer

The Storage Layer is where all data is stored.

Whenever data is loaded into Snowflake, it is automatically stored and managed in this layer.

One thing I found interesting is that Snowflake supports both structured and semi-structured data without requiring complex configurations.

The Storage Layer provides:

• Centralized data storage
• Automatic compression
• High durability and availability
• Support for structured and semi-structured data

Examples of supported data include:

• Tables
• CSV files
• JSON
• Avro
• XML
• Parquet

A simple way to think about it is this:

Imagine a large digital warehouse where all company data is stored safely.

No matter how many teams access the data, the information remains in one centralized location.

Compute Layer

The Compute Layer is responsible for processing queries.

This is where Snowflake’s Virtual Warehouses come into play.

Whenever a user runs a query, Snowflake uses compute resources from a Virtual Warehouse to process the request.

One of the biggest advantages of this layer is that it is completely independent of storage.

This means organizations can scale compute resources without affecting storage.

Virtual Warehouses can:

• Run queries
• Process ETL workloads
• Support dashboards and reports
• Scale independently
• Handle multiple workloads simultaneously

For example:

An analytics team may use one Virtual Warehouse for reporting.

A data engineering team may use another warehouse for data pipelines.

A data science team may use a third warehouse for experiments.

All three teams can work at the same time while accessing the same underlying data.

This was the point where Snowflake’s architecture started making a lot more sense to me.

Cloud Services Layer

The Cloud Services Layer acts as the control center of Snowflake.

Unlike the Compute Layer, it does not process queries directly.

Unlike the Storage Layer, it does not store business data.

Instead, it manages everything happening behind the scenes.

Some of its responsibilities include:

• Authentication
• Security
• Metadata management
• Query optimization
• Access control
• Governance

A simple analogy helped me understand this layer.

If the Storage Layer is a warehouse and the Compute Layer is the workforce, then the Cloud Services Layer is the management team coordinating everything.

It ensures the right users have access, queries are optimized, and resources are used efficiently.

How the Three Layers Work Together

Now let’s see what happens when a user runs a query.

The process looks something like this:

1. A user submits a query.
2. Cloud Services authenticates the user and checks permissions.
3. Query optimization takes place.
4. A Virtual Warehouse processes the query.
5. Data is retrieved from the Storage Layer.
6. Results are returned to the user.

Although this process happens in seconds, multiple components are working together behind the scenes.

Understanding this flow helped me appreciate why Snowflake performs so well at scale.

Why This Architecture Matters

At first glance, having three separate layers may seem like an implementation detail.

But this design provides several advantages.

Independent Scaling

Storage and compute can scale separately.

Better Performance

Different workloads can use different Virtual Warehouses.

Cost Optimization

Organizations pay only for the compute resources they actually use.

Workload Isolation

Multiple teams can work simultaneously without constantly impacting each other’s performance.

Reduced Maintenance

Snowflake manages much of the infrastructure automatically.

These benefits become increasingly important as organizations handle larger volumes of data.

Key Takeaways

Here are the biggest lessons I learned while studying Snowflake Architecture:

✅ Snowflake uses a three-layer architecture.

✅ Storage and compute are completely separated.

✅ Virtual Warehouses process workloads independently.

✅ Cloud Services manages security, metadata, and query optimization.

✅ Multiple teams can access the same data simultaneously.

✅ The architecture helps improve scalability, performance, and cost efficiency.

Final Thoughts

Understanding Snowflake Architecture was one of the biggest turning points in my learning journey.

Before learning about the three layers, many Snowflake concepts felt disconnected.

Once I understood the responsibilities of the Storage Layer, Compute Layer, and Cloud Services Layer, everything started making more sense.

If you’re beginning your Snowflake journey, I would strongly recommend spending time understanding the architecture first.

It creates a solid foundation for almost every other Snowflake concept you’ll learn later.