Apache Atlas is the open-source metadata and governance framework born in the Hadoop ecosystem. It provides a catalog, data lineage, and classification with a flexible type system, and it integrates tightly with Hive, HBase, Kafka, and the broader Hadoop stack. For on-prem Hadoop platforms, Atlas has been the standard governance layer.

Type system and entities

Atlas models metadata through an extensible type system: entities (a Hive table, a column, a process) and their relationships are defined as types you can extend with custom attributes. This lets it capture not just technical schemas but business context and custom asset types, all queryable through a metadata graph and search.

Lineage and classification

Atlas captures lineage automatically as data moves through Hive and Spark, so you can trace a table back to its sources and forward to its consumers. Its classification system tags data (e.g., "PII") and - integrated with Apache Ranger - can drive tag-based access policies: classify a column once and have access rules follow the tag rather than naming every column. That Atlas-plus-Ranger pairing is the classic Hadoop governance combination.

Where it fits

• On-prem Hadoop platforms (Cloudera/Hortonworks lineage) needing metadata, lineage, and classification.
• Tag-based access control via Atlas classifications + Apache Ranger.
• As a metadata backend - some other catalogs (including early Amundsen deployments) can use Atlas underneath.

How it compares

Atlas is powerful within the Hadoop world but is tied to that ecosystem and shows its age against cloud-native and modern-stack catalogs. For teams on cloud warehouses and the modern data stack, newer open-source options like DataHub and OpenMetadata, or commercial platforms, fit better and integrate more broadly. Atlas remains relevant primarily where an existing Hadoop estate already relies on it for governance.