Overall Satisfaction with Snowflake
My company adopted Snowflake as our first cloud-based data warehouse. It is being used as a central repository for all company data from each business unit for the purposes of business intelligence.
- Ease of use
- Separation of storage and compute resources
- Simple to scale up or down with virtual warehouses
- Built-in support for the most popular data formats
- Standard SQL dialect
- Robust function library
- Lacks support for common table expressions
- Lacks support for correlated subqueries
- Better technical support for customer identified bugs
- Clearer pricing model
- Centralized disparate data sources across the company
- Used to associate fragmented and unlock key insights
- Integrates easily with our business intelligence tool, Sisense
The average percentage of time that a data warehouse is actually doing something is around 20%. Given this, the price by query estimate becomes an important pricing consideration.
For this, Snowflake crucially decouples of storage and compute. With Snowflake you pay for 1) storage space used and 2) amount of time spent querying data. Snowflake also has a notion of a “logical warehouse” which is the “compute” aspect of the database. These warehouses can be scaled up or down to deliver different grades of performance. You can also configure the number of compute nodes to parallelize query execution. These warehouses can be configured to “pause” when you’re not using them for cost efficiency. As a result, you can have a super beefy warehouse for BI queries that’s only running when people are using your BI tools, while your background batch jobs can use cheaper hardware.
For this, Snowflake crucially decouples of storage and compute. With Snowflake you pay for 1) storage space used and 2) amount of time spent querying data. Snowflake also has a notion of a “logical warehouse” which is the “compute” aspect of the database. These warehouses can be scaled up or down to deliver different grades of performance. You can also configure the number of compute nodes to parallelize query execution. These warehouses can be configured to “pause” when you’re not using them for cost efficiency. As a result, you can have a super beefy warehouse for BI queries that’s only running when people are using your BI tools, while your background batch jobs can use cheaper hardware.