Home /

Snowflake Architecture

Rating: 5.0Blog-star
Views: 77
by Anjaneyulu Naini
Last modified: August 26th 2021

Snowflake was created in 2012 as a cloud-based data warehouse by 3 data warehousing professionals. Snowflake is a SaaS platform built on the top of Amazon Web Services (AWS) for loading, analyzing, and reporting on enormous data volumes.  Unlike typical on-premise systems that need hardware deployment, snowflake can be implemented in the cloud in minutes and is priced on a pay-per-second basis.

This article will assist you in developing a thorough knowledge of the Snowflake architecture, the way it holds and maintains data, and the ideas underlying its micro-partitioning. By the end of this article, you will also learn how Snowflake differs from the other cloud-based data warehouses..

Learn Snowflake Architecture

Points covered in this article are:

👉 What is Snowflake Data Warehouse?

👉 Distinguishing features of Snowflake Data Warehouse

👉 Snowflake Architecture

👉 Snowflake Architecture – A hybrid model

What is Snowflake Data Warehouse?

Snowflake is one of the only cloud-based data warehouse solutions that prioritize simplicity over functionality. It automatically scales up and down to achieve the optimal performance/cost ratio.

With Snowflake, you can centrally store all of your data and scale your computing independently. For instance, if you require heavy data loads for complicated transitions but only have a few significant queries in your reports, you may create a large Snowflake warehouse for the data load and then scale it back down once complete – all of this in real-time. This reduces costs without affecting your objectives.

Distinguishing Features of Snowflake Data Warehouse

1. Cloud Agnostic Solution

Snowflake is a professional data warehouse solution that runs on all three major cloud providers: AWS, Google Cloud Platform, and Azure all with the same consumer experience. Customers can simply integrate Snowflake into their existing cloud infrastructure and deploy it in systems that make commercial sense.

2. Scalability

Snowflake enables customers to optimize resources when huge volumes of data need to be uploaded quickly and down again when the operation is terminated without affecting service. Customers can begin with an extremely small cloud warehouse and grow up or down as necessary. Snowflake includes auto-scaling and auto-suspend capabilities to ensure minimal management.

3. Separation of Concurrency and Volume of work

Customers would thrive for resources in a typical data warehouse system, resulting in concurrency difficulties. Synchronization is no longer a problem due to Snowflake's multi-cluster design. One of the primary advantages of this design is that it allows for the separation of workloads to be run against their own computing clusters, referred to as virtual warehouses. Queries executed against one cloud warehouse would never have an effect on queries executed against another.

4. Security

Snowflake incorporates a variety of protective measures, from the means consumers use the platform to the way data is kept. You may control network policies by whitelisting IP addresses that you want to prevent from logging into your account. Snowflake supports a variety of authentication techniques, like two-factor identification and federated authentication for single sign-on.

Want to enhance your skills to become a master in Snowflake Cloud Data Warehouse, Enroll in our Snowflake Online Course

Snowflake Architecture

Snowflake architecture is a mix of shared-disk and shared-nothing structures that combines the advantages of each. Let us explore each of these designs and see how Snowflake integrates them to create a new hybrid architecture:

1. Shared-disk architecture: It is commonly used in conventional databases and consists of a single storage layer that is available to all grade levels. Multiple cluster nodes equipped with CPU and memory connect with the centralized storage layer to retrieve and interpret data.

shared disk architecture

2. Shared-nothing architecture: Unlike the Shared-Disk design, it utilizes dispersed cluster nodes that each have their own disc storage, CPU, and memory. The benefit is that because each cluster node has its own storage space, data could be divided and saved among these cluster nodes.

Snowflake shared nothing architecture

Snowflake Architecture – A Hybrid Model

A snowflake is composed of three distinct layers:

#1 Storage Layer

Snowflake divides the information into many internal optimized and compressed micro partitions. It stores data in a columnar fashion. Data is saved in the cloud and is managed using a shared-disk architecture, which simplifies data administration. This ensures that customers in the shared-nothing paradigm are not concerned about data transmission across many nodes.

Computer units establish connections to the storage layer in order to retrieve information for query processing. Users just spend for the monthly average storage usage because the storage layer is self-contained. Because Snowflake is cloud-based, storage space is elastic and paid monthly based on consumption per TB.

#2 Compute Layer

Snowflake executes queries using a "Virtual Warehouse". Snowflake maintains a layer of separation between the query processing layer and the disc storage layer. This layer executes queries against the data in the storage layer.

Virtual Warehouses are computing units consisting of several nodes with Snowflake-provisioned CPU and Memory. Snowflake allows for the creation of several Virtual Warehouses to meet a variety of needs depending on the workload. Each virtual warehouse may be configured to use a single storage tier. In general, a virtual warehouse operates independently of other virtual warehouses and does not communicate with them.

#3 Cloud Services Layer

This layer contains all the operations that coordinate throughout Snowflake, like authorization, encryption, metadata for loaded data, and query processing. Types of services handled by this layer include the following:

  • Whenever a login process is initiated, it must traverse this layer.
  • Snowflake queries are routed through this layer's analyzer and later to the Compute Layer for execution.
  • This layer stores the metadata necessary to improve a query or filter data.

snowflake three layer architecture

All three layers are self-scaling, and Snowflake bills separately for disk and virtual warehouse. The services layer is managed inside provisioned computing nodes, and so is not priced. The benefit of the Snowflake design is that each layer may be scaled independently of others.

Conclusion

Snowflake comes with a slew of features pre-installed. A simple-to-use platform, like Snowflake, may go a long way in improving your data warehouse use cases making it simpler to create and sustain. We hope this blog helped you gain a deeper insight into Snowflake Architecture. 

About Author

author
NameAnjaneyulu Naini
Author Bio

Anjaneyulu Naini is working as a Content contributor for Mindmajix. He has a great understanding of today’s technology and statistical analysis environment, which includes key aspects such as analysis of variance and software,. He is well aware of various technologies such as Python, Artificial Intelligence, Oracle, Business Intelligence, Altrex etc, Connect with him on LinkedIn and Twitter.