One of the many new feature announcements at this year’s Snowflake Summit was Snowflake Unistore Hybrid Tables.
The Hybrid Table offering stores data to support fast single-row transactions along with storing the data in columnar format. In other words:
- The data is stored to support both operational and analytical workloads
An industry acronym often used to describe this approach is HTAP : Hybrid Transactional and Analytical Processing. While the acronym may have been coined by Gartner in 2014, and referred to as “Translytical” by Forrester Research, various academic research predates that.
Snowflake Hybrid Tables in a nutshell:
- Optimized to support both operational workloads e.g. single row operations, and analytically workloads on the same platform
- As mentioned this optimization comes from storing the data for operational purposes, but also in columnar format which is Snowflakes ‘bread & butter’. The columnar format is compressed which means less disk I/O and lends itself really well to aggregational computation from a CPU perspective
- Primary & foreign key enforcement – Primary key is required
- Supports unique & non-unique indexes
- Row level locking & read-committed isolation level
- Snowflake will determine automatically whether to run an OLTP or Analytical query on a Hybrid Table
- Hybrid Tables are the tech behind the Unistore workload. As described by Snowflake: “Unistore delivers a modern approach to working with transactional and analytical data together in a single platform“
Hybrid Tables are in private preview at present, so apart from a small audience they are not readily available to test run just yet. The billing aspect has also yet to be announced, and this will be important in terms of configuration decision making.
Depending on the use case, Hybrid Tables will offer a practical approach to working with both operational & analytical workloads while not having to move data. This means also benefiting from the Snowflake Data Cloud components such as elasticity, governance, programmability & collaboration.
© Dan Galavan. Dan is an independent data architecture consultant working with data for 23 years. He is a Snowflake Data Cloud specialist, data modeling expert, and a member of the Snowflake Data Super Hero program. Get in touch.