SlateDB
An open-source embedded LSM-tree database that uses object storage for durability, offering high availability, low latency, and cost efficiency without local disks.
SlateDB is an open-source alternative to:
SlateDB is an open-source embedded key-value database built on an LSM tree architecture that relies solely on object storage for durability. Designed for modern cloud-native applications, it eliminates the need for local disks while maintaining 99.999999999% durability and transactional semantics.
Key Features
- Zero-disk architecture: Persist data directly to object storage (S3, R2, GCS) without local disks or cross-AZ traffic, reducing costs and complexity.
- Flexible deployment: Single-writer multi-reader model with formally verified manifest fencing for consistency, supporting embedded or separate compaction processes.
- Transactional semantics: Full ACID transactions, snapshots, and optimistic concurrency control for demanding OLTP workloads.
- Tunable performance: Optimize write durability latency vs. API request costs and leverage hybrid memory/disk caching for reads.
- Clones & branches: Zero-copy checkpointing enables efficient cloning, forking, and rescaling operations with O(1) splits and merges.
- Multi-language support: Rust core with Go, Java, Node.js, and Python bindings for broad compatibility.
- Cloud-agnostic: Runs on AWS S3, Cloudflare R2, Google Cloud Storage, and most S3-compatible object stores.
- Apache 2.0 license: Free and open-source with permissive licensing.
Tech / Self-Hosting
SlateDB is designed for self-hosting and cloud-native environments. It requires no local storage infrastructure and can be deployed as an embedded library or with separate compaction processes. The system is optimized for object storage performance and cost efficiency, making it ideal for scalable, distributed applications. Configuration is handled through standard object storage settings, and the project provides comprehensive documentation for quick setup and tuning.