Databricks guide argues real-time analytics belongs inside the lakehouse

Databricks says real-time analytics should run as a feature of its lakehouse rather than on a separate serving engine such as Apache Druid, Apache Pinot, or ClickHouse.

The guide, published on the Databricks Blog, defines real-time analytics across three dimensions: freshness, latency, and concurrency.

The guide separates real-time workloads, which need insights in milliseconds to seconds, from batch analytics, which can wait minutes to hours. The source is first-party, so these claims are Databricks' own position on the architecture.