Snowflake: Unleash the potential of an innovative architecture

11
min
Created in:
Jun 10, 2024
Updated:
9/27/2024

Snowflake is more than just a data cloud platform; it combines a native hybrid architecture with the advantages of traditional database architectures.

We know that with the rapidly increasing pace of new technologies, it’s becoming increasingly difficult to keep up with the emerging data platforms like Snowflake and even more challenging to choose the best option.

Therefore, it is of utmost importance to be familiar with new and more modern tools to stay updated with the market and obtain the best data-driven solutions.

And Indicium is here to help you in this process!

This edition focuses on a tool that has been receiving significant attention: Snowflake.

We will analyze how this data platform redefines the concept of data storage and analysis, bringing the advantages of traditional architectures and elevating them to a new level of efficiency and innovation.

Happy reading!

What is the basis of Snowflake's architecture?

Before looking at Snowflake’s architecture, it’s important to understand some concepts.

Its architecture is essentially a hybrid of distributed computing architectures, combining shared-disk and shared-nothing database architectures.

When you have a shared-disk structure, the same disk or storage device is shared.

Thus, although each node has its own memory, they have access to the same data from replicas present in each node.

Due to the shared access, it is necessary to monitor and control data processing through cluster control software.

Representation of Snowflake's cluster control software. There are three squares labeled CPU, categorized as clusters, which are directed to a cylinder labeled storage.
Diagram of Snowflake's Shared Disk Architecture.

When thinking of a shared-nothing database structure, imagine that each node has its own memory and is independent.

In other words, they do not share storage or disk space and are interconnected through a network.

Illustration of a diagram of Snowflake's shared-nothing architecture. There are four pairs of clusters, each represented by a CPU and storage, interconnected to a cloud labeled network, positioned in the center.
Diagram of Shared-Nothing Architecture.

To understand which architecture would best meet your needs, it is necessary to analyze the advantages and disadvantages of each type of distributed architecture.

Below, we have listed some of the pros and cons of each type.

Two tables listing the advantages and disadvantages of Snowflake's architectures, Shared Disk and Shared Nothing, with the latter having more advantages and fewer disadvantages.
Advantages and disadvantages of Snowflake's Shared Disk and Shared Nothing architectures.

Snowflake: The best of both worlds with a multi-cluster architecture

Most traditional data storage and analysis systems organize their hardware into one of the two distributed computing architectures, Shared Disk or Shared Nothing.

Snowflake consists of a service-oriented architecture composed of three physically separate but logically integrated layers.

Refer to the image below:

Illustration of Snowflake's architecture, which has three separate layers: cloud services, query processing, and database storage.
Snowflake Architecture

The first layer, Cloud Services, is a set of services that manage activities in Snowflake, such as processing user requests.

The services included are:

  • Authentication
  • Infrastructure management
  • Metadata management
  • Query analysis and optimization
  • Access control

The second layer, Query Processing, is composed of separate computing clusters called virtual warehouses, which are responsible for executing the computation needed to process a query.

They use SQL commands to create the warehouses managed by Snowflake.

This is where the clusters operate similarly to the Shared Nothing architecture.

The last layer is Database Storage, a centralized cloud storage layer that holds all the data available in databases, similar to the Shared Disk architecture.

Snowflake Architecture vs. Standard Warehouses

Snowflake's multi-cluster architecture has advantages compared to standard warehouses.

For example, there is no need to manually increase the size of the warehouse, start additional ones, or reduce its size.

With its two modes, maximized and auto-scaling, it is possible to connect a larger number of users to the same warehouse size.

When using the maximized mode, you can control the capacity of the multi-cluster, upgrading or downgrading the number of clusters as needed.

On the other hand, in auto-scaling mode, Snowflake automatically starts and stops additional clusters as needed.

It does this without resizing, pausing, or starting the warehouse, or stopping additional ones to handle variations in workloads.

How can Snowflake help your company?

Snowflake is more than a data cloud platform; it combines native hybrid architecture with the advantages of traditional database architectures.

It is also a high-performance platform by separating the scaling of computing resources from storage. Additionally, it can help your company by:

  • Automatically scaling computing resources
  • Transparently provisioning resources
  • Automatically managing metadata
  • Customizing the computing engine for each workload
  • Querying semi-structured data relationally
  • Securely sharing data within and outside your organization

Thus, using Snowflake's multi-cluster architecture improves the scaling of resources for concurrent user and query usage, providing more autonomy in managing projects.

Want to know more about what Snowflake can offer for your company or the one you work for?

Contact Indicium.

We are the only Brazilian Select Snowflake partner.

This means we are certified, partners of the platform, and have already implemented it in several companies in Brazil and worldwide.

Click here to talk to our team.

Tags:
Snowflake
Analytics
All
Data analytics
Tools
Partnerships

Emy Kuroiwa

Engenheira de analytics

Keep up to date with what's happening at Indicium by following our networks:

Prepare your organization for decades of data-driven innovation.

Connect with us to learn how we can help.