SlateDB is an embedded key-value storage engine that writes directly to object storage. It uses a log-structured merge (LSM) tree design, but instead of putting its files on a local disk, it puts them in object storage such as S3, GCS, Azure Blob, or MinIO. It is a library you build into your own system rather than a database you connect to over the network.
Why it is here
SlateDB is a building block: a library you embed, not a managed database an application queries directly. It is the storage layer you would reach for to build a lakebase of your own. Because it treats object storage as the primary store, it inherits the same properties that define the category:
- Object storage is the disk. Durability lives in object storage, so any process can be torn down and recreated without losing data.
- Zero-copy cloning. Copy-on-write semantics allow instant clones and branches, the same trick that makes branching cheap in a relational lakebase.
- Separated compute. Any number of readers can attach to the same durable data.
Where it sits in the category
SlateDB is a building block rather than a finished lakebase. It shows that the lakebase architecture is a general pattern, not a single product. When you see a new operational database that claims instant branching and scale to zero, there is often a component like SlateDB underneath, turning object storage into a durable, cloneable disk.