Snowflake Data Cloud now enables Native Apps


This is a new strategic shift for American publishers. Its offering is no longer just a data platform, but an application -focused platform as a service environment.

On the occasion of the Snowflake Summit, held this week in Las Vegas, the American publisher of the same name is announcing a new phase in its strategy. Combining the flexibility and volume of big data with the performance of a data warehouse, its lakehouse now makes it possible in cloud mode (Data Cloud) to propel applications. Until now limited to exchanging data from its platform, the publisher’s marketplace also provides a place for them to share and monetize them.

“Based on our environment, these data apps will benefit from transactional or analytical data, but also embed machine learning models as needed,” explains Benoît Dageville, co-founder and product president of Snowflake. The centerpiece of the device, the Native Application Framework, which is currently in private beta, manages their entire life cycle, from development to sales through deployment and scaling up. Applications can clearly take advantage of Snowflake -specific functions: stored procedures, UDF, UDTF …

Downstream, apps built on Snowflake and associated data remain in the space (or tenant) of the client to maneuver. The latter will remain in control of their security and management regardless of their use. Of Snowflake’s 425 partner publishers, some already have access to the new framework. This is the case of Informatica or ServiceNow as part of the development of new Snowflake connectors. But also from Google to integrate Google Analytics indicators into the platform.

Machine learning at the heart of the offer

The Native Application Framework fits Snowpark. A library designed to process data in a massive parallel manner in Snowflake (in the Spark model) while taking advantage of a sandbox to secure it. Its purpose? Enable the generation of data pipelines and machine learning processes. During its global event, Snowflake also announced the availability of Snowpark’s public beta for Python. The language has been added to Java and Scala, languages ​​already supported. Snowpark for Python integrates with the Python development environment resulting from the acquisition of Streamlit. At the same time, many Python libraries are supported: Numpy and Pandas on the data analytics side and Scikit-learn and Tensorflow on the machine learning side.

Snowflake cloud architecture. © Snowflake

To complete Snowpark for Python, Snowflake is working on several developments. First, Snowflake Worksheets for Python which aims to integrate Streamlit into its core GUI (Snowsight). Then, SQL Machine Learning to be adapted to build predictive machine learning models based on time series. These two bricks are currently in private beta. Finally, the Large Memory Warehouse, which is currently under construction, will perform memory-intensive operations, such as feature engineering or machine learning processing applied to large data sets.

Transactional and data analytics

Along with the introduction of the apps, Snowflake Summit saw other big announcements. Main of which is the introduction of UniStore. Launched in private beta, this brick powers Snowflake’s transactional processing management. Stated goal: to run data services with latencies of a few milliseconds by controlling states and concurrent access. “That’s very useful in machine learning”, argues Benoît Dageville. For the occasion, Snowflake makes hybrid tables to manage both transactional processing (OLTP) and data analytics. “The goal is to avoid sharing between current data and historical data with the need to have ETL between the two”, added the Snowflake co-founder.

Another announcement, Snowflake improves real-time data ingestion. An evolution that includes the launch of Snowpipe Streaming (in private beta). A technology that allows data to be streamed in serverless mode. To this brick will soon be added Materialized Tables, a feature being done that aims to simplify the declarative transformation of data. To interface with third-party databases, Snowflake also works with two new types of tables. In development, the first, Iceberg Tables, will open a gateway to the Apache Iceberg table format. The second, External Tables for On-Premises Storage, will provide access from Snowflake to storage systems deployed internally, such as Dell Technologies and Pure Storage.

When it went public in September 2020, Snowflake raised $ 3.4 billion. The company has more than 3,000 employees for more than 6,300 customers worldwide.

Leave a Comment