Why Snowflake changed the rules of data warehousing
When Snowflake arrived on the market, it introduced an idea that seems obvious today but was revolutionary at the time: fully separating storage from compute. In traditional data warehouses, if you needed more processing power, you also had to buy more storage, and vice versa. With Snowflake, each scales independently, and compute can be turned on and off in seconds.
This architecture fundamentally changes the economics of data warehousing: you pay for what you use, not for the total capacity you’ve provisioned.
Snowflake architecture: the key concepts
Compute-storage separation: data is stored in compressed columnar format in cloud object storage (S3, Azure Blob, or GCS, depending on your cloud). Virtual warehouses are independent compute clusters that read that data without interfering with one another.
Virtual warehouses: each team or workload can have its own right-sized virtual warehouse. Marketing uses an XS for its routine queries, data science uses an L for heavy transformations, and both work on the same data without blocking each other.
Multi-cloud and multi-region: Snowflake runs on AWS, Azure, and GCP, and can replicate data across regions and clouds for redundancy and regulatory compliance. The architecture is the same regardless of the underlying cloud.
Time Travel: Snowflake lets you query the state of your data at any point within a configurable period (up to 90 days on Enterprise editions), which makes it easier to recover from errors and run historical analysis.
Snowflake Data Marketplace
One of Snowflake’s most distinctive features is its Data Marketplace: an ecosystem of third-party data that companies can bring straight into their Snowflake account without moving or processing it. Weather data, demographics, financial indices, macroeconomic data, and thousands of other datasets are available to enrich your own data with no extra infrastructure.
For companies in insurance, finance, or retail, access to high-quality external data can be a significant competitive edge.
Snowpark: ML directly in Snowflake
Snowpark is Snowflake’s API that lets data scientists run Python, Java, or Scala code directly inside Snowflake, using Snowflake’s compute without extracting data to an external environment. This cuts data movement, simplifies ML architecture, and makes it easier to deploy models directly in Snowflake.
Snowflake vs. Databricks vs. BigQuery: when to choose each
Snowflake is the best choice when:
- Your core team is SQL analysts and BI teams, not Spark engineers.
- You need to share data securely with external partners or suppliers.
- You want a managed platform with minimal operational overhead.
- You operate across multiple clouds and need interoperability.
Databricks is preferable when:
- ML and data engineering workloads outweigh analytical SQL.
- You process massive volumes of unstructured data.
- Your team has Spark and Python experience.
BigQuery is the natural fit when:
- You’re already deeply embedded in the Google Cloud ecosystem.
Implementation for businesses
Snowflake has a relatively gentle learning curve for teams with SQL experience. A typical implementation covers designing the data model in Snowflake, setting up ingestion pipelines (with tools like dbt for transformations), defining the virtual warehouse policy for each team, and integrating with existing BI tools such as Power BI or Tableau.
At AIRES, we help companies assess whether Snowflake is the right platform for their needs, design the best architecture, and manage migration from existing systems safely.
Frequently asked questions
How do you control costs in Snowflake? Snowflake bills for compute credits consumed and for storage. You control costs with auto-suspend policies for virtual warehouses, right-sizing each warehouse for its use case, clustering keys on large tables to reduce the data scanned, and usage monitoring through Snowflake’s Query History views.
Can you use Snowflake with Power BI? Yes. Power BI has a native Snowflake connector that links reports directly to Snowflake tables and views, in both Import and DirectQuery mode. This setup is very common among companies that want to keep Power BI as their BI tool while modernizing their data back end.
Is Snowflake GDPR-compliant for European companies? Yes. Snowflake offers EU regions on AWS Frankfurt and Azure Netherlands, among others, and data can be configured to stay within the EEA. Snowflake provides SOC 2 Type II and ISO 27001 certifications and GDPR compliance, including data processing agreements (DPAs) tailored to European regulation.