A storage engine delivers scalable analytic performance and capacity for self-managed, on-premises or cloud deployments, by reading only the columns a query needs.

A shared-nothing compute and shared-object storage architecture lets you scale compute resources independently of storage.

An MPP-based query engine runs analytical queries across all nodes in parallel, for scalable, real-time results.

A cost-based optimizer chooses efficient query plans using statistics gathered by ANALYZE COLLECTION.

Stream data in real time from Couchbase Server, Couchbase Capella, and nearly any other database or application through Kafka. Analyze incoming data in its native JSON form, with no flattening or transformation pipelines, and optionally modify the target JSON structure while data is in transit.

Query data directly in supported object storage, in JSON, Parquet, Avro, CSV, TSV, Delta table, and Apache Iceberg formats, with Iceberg tables accessed through Iceberg catalogs. A single query can combine external data with data already in Operational Insights.

Use SQL++ to write query results back to the Couchbase operational data service, to support adaptive applications.

Tabular views give Tableau, Power BI, and Apache Superset native SQL-based access, for building business reports, visualizations, and dashboards.

Create a vector index for approximate nearest neighbor (ANN) search over embeddings, and use the CLUSTER BY clause to group embeddings by similarity.

Extend SQL with Python functions running in an isolated sandbox, to bring a trained ML model into a query or implement logic that's hard to express in SQL.

Operational Insights reuses a cached execution plan for a repeated query, skipping the optimization step to start executing faster.