Publishing features to the Databricks Online Store

If you want to use your features with real-time serving, you can publish the features to a low-latency database, also known as an online store. Publishing feature tables to a low-latency database allows for automatic feature lookup during model inference. There are a variety of options to choose from when choosing an online store. A typical data-serving solution requires expert engineers to select an appropriate database for online access, build data publishing pipelines, and orchestrate deployment. After deployment, someone has to monitor, manage, and optimize the pipelines feeding the online store. This is why we recommend Databricks’ own fully managed serverless online store built right into the platform. It automatically syncs your Delta feature table with the online store, making it amazingly easy to use. Databricks Online Store is integrated with Databricks Model Serving, so it’s easy to set up your online store...