By David Teplow, Managing Director of Snowflake Services
It was an exciting time to be at Snowflake Summit (June 13-16, 2022). It was the first non-virtual Summit in three years and in that time Summit attendance, with nearly 9,000 in-person attendees, grew by more than an order of magnitude; as did Snowflake’s annual revenues (see image below).
The most exciting thing for me, personally, was to actually spend time with so many of the people I worked closely with for over a year as part of Snowflake Professional Services but never got to meet in-person, including some of the Solutions Architects and Service Delivery Managers I hired and managed as part of the Northeast PS team.
There were exciting keynotes and presentations, as well. In my view, two of the most compelling and significant announcements at Summit were support for Hybrid Tables (a.k.a. Unistore) and Native Applications. Combined with the announcement that the Snowflake Data Marketplace will now simply be called the “Snowflake Marketplace”, tells me that Snowflake is all in on SHARING. Smart move; because Secure Data Sharing is a distinct competitive advantage Snowflake has over its competition. It’s apparent to me that they’re pressing that advantage by expanding what can be shared from just data to the applications built around that data. To do that effectively, you not only need the data but also ways to manipulate that data both analytically and transactionally (hello Hybrid Tables), as well as an application framework that operates across Snowflake accounts and across cloud providers / regions (hello Native Applications).
Delving into these announcements a little deeper, Unistore opens up the world of Transactions (OLTP) to Snowflake, which until now has been focused on Analytics (OLAP). Analytical workloads on large datasets benefit greatly when data is stored in Columnar format (pioneered by Vertica in 2005; subsequently used by Redshift, Synapse, BigQuery and Snowflake). Transactional workloads, on the other hand, work best using traditional row-based storage. Hybrid Tables (currently in private preview) are the new Snowflake table type behind Unistore. They use row-based storage and support unique primary keys (which are enforced in Hybrid Tables as opposed to merely documented in normal Snowflake tables), referential integrity constraints and row-level locking – essential ingredients for reliable and efficient transaction processing. To support Snowflake’s bread-and-butter analytical processing, data in Hybrid Tables is replicated from its row-based storage to native columnar storage (see image below).
More details on Unistore can be found at: https://www.snowflake.com/blog/introducing-unistore/
The announcement of Snowflake’s Native Application Framework (currently in private preview) was significant because it paves the way for users to share not only their data but also the code they’ve developed to refine, contextualize and turn raw data into insightful information and actionable knowledge. Along with sharing Tables and Views, you can share an Installer Package, Stored Procedures, UDFs and even Streamlit integration (currently in development) as a fully-functioning application (see image below).
The framework even includes telemetry tools that allow you to monitor and support your applications. More details on Snowflake Native Application Framework can be found at: https://www.snowflake.com/blog/introducing-snowflake-native-application-framework/
What you’ll start to see on the Snowflake (not just Data) Marketplace is more than just datasets that can be woven into your internal data, but code sets – including complex calculations and machine learning algorithms – that can be woven into your internal applications. This makes the Marketplace all the more valuable, which in turn makes Snowflake all the more valuable.