While we think Snowflake is a great cloud data warehouse solution, the optimal cloud data warehouse for you is entirely based upon your business use-cases and stakeholder needs. We also recognize that the pricing and functionality of the warehouse should not be the sole determining factors in your data warehouse decision. As we have already stated, we highly recommend clients and data warehouse shoppers to consider Snowflake as we think it has the broadest range of use cases. Below are three reasons why Untitled prefers Snowflake and why it might be a great fit for your business.
1. Transparent Pricing
Snowflake provides very transparent pricing. With its simple pricing structure, it’s easy to calculate what your costs would be using the platform and even build a POC. If you’re a Redshift user, you can look through your telemetry data and determine how often you use each cluster. This should give you a solid number to build out a pricing model if you were to convert from Redshift to Snowflake. If you’re a BigQuery user, you could view the number of queries run in your warehouse and determine the average query time and use that to determine that amount of time your warehouse in Snowflake would be running. At Untitled, we find that Snowflake beats Redshift and BigQuery on cost at least 80% of the time.
2. Integrates Seamlessly With All Major Cloud & ETL Providers
As has already been mentioned, Snowflake integrates seamlessly with all the major cloud providers. Whether you are using Azure, GCP, or AWS, your underlying storage can be located within either provider and all of your data can flow seamlessly into the tools you use within those platforms. On top of that, all of the major ETL and business intelligence tools often prefer to use Snowflake as the primary destination for your data. This means that your experience using these tools is often more robust than if you were using a different warehouse. An example of this is the relationship that Fivetran, dbt, and Snowflake have in providing highly valuable data extraction and transformation packages available for their customers.
3. The Ability to Split Up Warehouses
Finally, because the warehouse (compute) is the foundation of how Snowflake works, you can have an environment where warehouses can be split up to serve your business use cases. This also allows for you to have an extremely granular understanding of where your costs are coming from, as well as bottlenecks in your overall data warehouse architecture. For example, a typical best practice that we have seen when setting up Snowflake is to have three warehouses; a data loading warehouse, transformation warehouse, and an analytics warehouse.
The loading warehouse is reserved exclusively for data loading from custom ETL or third-party tools like Fivetran or Stitch. The transformation warehouse is reserved for transforming the data from the raw format that it flows in as, into useful data models and tables. This often includes dbt and an orchestration tool like Airflow if you are using a tool outside of dbt to run your jobs. Finally, the analytics warehouse is used for just that, analytics. Whether that is an analyst running queries, or a business intelligence tool like Tableau, this warehouse allows you to understand what the consumption of that data is.
To reiterate, this setup allows for you to see where costs are occurring within your warehouse, which processes are running the most, and an understanding of how these warehouses are being utilized from an under or over-utilization standpoint. We believe that this can also have compounding positive effects as it may bring light into inefficiencies that are running elsewhere in your data stack.
Cloud data warehouses are not going away any time soon. While arguments can be made that a data lake architecture is more cost-effective, there is undeniable flexibility and functionality that comes from using a cloud data warehouse. We are also keenly aware that Redshift, BigQuery, and Snowflake are not the only options available.
In addition, we have been eyeing Firebolt very closely and think that this warehouse may even give Snowflake a run for its money. Because Untitled is cloud-agnostic, we feel like we understand the strengths and weaknesses of the major cloud warehouses. While this article is not exhaustive of all the factors that play into the final decision of choosing a cloud data warehouse, we would love for you to reach out if you have additional questions.
If a cloud data warehouse is something you need, or you need a partner to help you implement, optimize, or help advise you on choosing one, we would love to help make that happen.