HydraDB is a distributed graph database written in Rust that stores data on object storage and executes queries on disaggregated compute nodes.
The tool addresses the challenge of building scalable graph databases by separating storage from compute. Graph data lives durably in S3-compatible object storage, while two independent compute roles handle different workloads: data nodes serve queries and apply mutations, while indexers build traversal indexes asynchronously in the background. Both roles maintain only disposable state locally, allowing them to scale or be replaced without moving the underlying graph. Queries run against pinned snapshots for consistency, and indexed traversal combines compiled sparse formats with visible write-ahead log overlays to ensure correctness even when indexes lag behind mutations.
Teams should consider this tool if they need a graph database that scales compute independently from storage, want to avoid moving large datasets when adding capacity, or prefer to keep their graph in object storage they already operate. The project offers Neo4j-compatible Bolt connectivity and an HTTP API with typed JSON and streaming NDJSON responses, making it accessible to applications built around standard graph clients. It includes graph-native query execution using property indexes, reverse adjacency, and GraphBLAS sparse traversal where beneficial. The tool is suitable for workloads where disaggregated architecture and object-store durability outweigh the complexity of managing separate data and indexer roles.
Development activity shows consistent attention to core infrastructure and operational concerns. The project maintains active work on the storage layer, query execution pipeline, and writer coordination mechanisms that underpin the disaggregated architecture. There is ongoing development of observability features including authentication, authorization, metrics, and tracing. The codebase receives regular updates to both the data node and indexer components, indicating sustained focus on keeping both roles functional and performant.
