Best practices

Best practices guide for optimal performance when using Snowflake

Snowflake is a fully managed cloud data platform known for its innovative consumption-based pricing model and elastic scalability. By connecting Snowflake to Tellius, you can efficiently analyze large volumes of data without extensive infrastructure management. The guidance below helps ensure optimal performance and ease of use when working with Snowflake data in Tellius.

Live mode integration

Live Mode enables Tellius to execute queries directly against Snowflake, leveraging Snowflake’s powerful engine without extracting or copying data into Tellius’s internal storage.

• Queries always run on the most up-to-date Snowflake data, ensuring real-time accuracy.

• No data duplication—saves storage costs and reduces data latency since you’re not maintaining another copy of the data.

For most Snowflake users, enable Live Mode. This mode reduces ETL overhead, streamlines administration, and takes advantage of Snowflake’s native optimizations.

Indexing & Partitioning

• Although Snowflake does not use traditional indexes like legacy databases, it automatically optimizes data storage (micro-partitions) under the hood. To influence micro-partition pruning, consider creating appropriate indexes and clustering keys on frequently-filtered columns.

• Identify columns often used in filters, group by clauses, or join conditions. Defining clustering on these columns improves query pruning and accelerates performance.

Data Modeling

• Ensure that the data model in Snowflake is clean, well-structured and follows best practices (e.g., proper normalization, clear primary keys).

• Using a STAR or Snowflake schema simplifies queries and can reduce complexity when joined within Tellius.

Complex Views

Instead of building intricate calculations or views in Tellius, create them directly in Snowflake. Snowflake’s engine is optimized for handling complex SQL logic.

Transformations in Live Mode

• If you’re using Live Mode, all major data transformations—such as complex joins, aggregations, or window functions—should be performed in Snowflake.

• Within Tellius, limit changes to adding synonyms, aliases, or simple field-level customizations that do not require full data transformations. This keeps your analytics layer lean and agile, while the computationally intensive work remains on Snowflake.

Insights and Machine Learning (ML)

• Providing Tellius with detailed, granular, and less-aggregated data from Snowflake results in richer insights and more accurate ML models. The reason is that aggregated data can mask patterns and anomalies that matter for predictions and data discovery.

• If more than 40% of your Tellius usage involves running Insights and ML workloads, consider creating a secondary Snowflake connection:

  • Disable live mode to ensure this connection imports data into Tellius at scheduled intervals.

  • Running insights on a periodically refreshed, non-live dataset can increase performance and reduce query costs. You utilize Snowflake’s power when needed but rely on a cached dataset in Tellius for rapid exploratory analysis and ML model building.

Joining Snowflake data with external sources

• You can join Snowflake data with other data sources directly in Tellius.

• Follow best practices of warehouse design—such as STAR and Snowflake Schemas.

  Ensure the columns used for joining (keys) match in data type, distribution, and cleanliness. This step prevents inefficient queries and inconsistent results.