Stream as a storage primitive for the lakehouse paradigm
Ursa is a storage engine that implements Lakestream, an open API and specification for stream storage on object storage, with a stream materialization framework that defines how a stream becomes a lakehouse table. Ursa turns any object, block, or file store into stream storage, and it materializes streams into lakehouse tables and other queryable states.
Ursa 1.0 is published to Maven Central under the org.openlakestream group; the quickstart embeds it in a Java program. It is built to be embedded in messaging brokers: Ursa for Apache Kafka (UFK) is a Kafka distribution built on the Lakestream API and specification that stores its diskless topics through Ursa.
The storage is designed with the following principles.
Ursa retains stream records in WAL and Compacted Objects with an Offset Index for streaming reads and replay. Optional table materialization delivers records into external analytical tables whose files and lifecycle are independent of internal stream storage.
By turning any object, block, or file store into shared log or stream storage, Ursa enables diskless event-streaming integrations without coupling the storage engine to a broker protocol. This provides:
- No single point of failure or bottleneck for partitions
- Reduced cross-AZ network traffic
- Lower operational complexity (no leader elections or rebalancing)
- Rebalance-free architecture for event brokers
Background materialization delivers stream records to external tables using the same WAL read pass as internal compaction. Applications can consume retained records through the stream API and query the delivered copy through the destination table's APIs.
- Stream-First Storage Primitive: Unified abstraction for data streams and lakehouse tables
- Multi-Cloud Support: AWS S3, Google Cloud Storage, Azure Blob Storage
- Lakehouse Integration: Delta Lake and Apache Iceberg with multi-catalog support
- High Performance: Read/write caching, batching, and prefetching optimizations
- Observability: Metrics for the storage and compaction paths (see Metrics)
See Build & Run Locally for complete build and local run instructions.
See Contributing Guide for information on how to contribute to the project.
- Issues: Report bugs and feature requests on GitHub Issues
- Documentation: See docs/ for detailed guides
- Questions: Ask in GitHub Discussions
- Specification: The Lakestream API and specification are documented at openlakestream.org
Ursa is licensed under the Apache License, Version 2.0.
The binary distribution contains only Apache 2.0 and other permissively licensed third-party jars. JSON Schema and Protobuf records are decoded without the Confluent Community License provider artifacts; see Third-party license notes for details and the build guard rails.