Zilliz Cloud and Milvus
A vector database with explicit storage and compute separation.
Milvus is an open-source vector database, and Zilliz Cloud is the managed service built on it. Its architecture was designed around the separation of storage and compute from the start. Stateless worker components handle indexing and querying, while durable data is persisted to object storage. That design is a lakebase applied to vector search.
Why it qualifies
- Object storage for persistence. Vector data and indexes are persisted to object storage (S3 or MinIO), keeping durability independent of any single node.
- Storage and compute are separate. Milvus runs as a set of stateless microservices, often on Kubernetes, so query and indexing capacity scale independently of stored data. Hot and cold tiers keep frequently used vectors close to compute.
- Operational workload. It serves live similarity search for recommendation, retrieval, and AI applications.
Where it sits in the category
Like turbopuffer, Zilliz and Milvus serve vector search, an operational workload, rather than transactions. They fit the lakebase category under the broader reading: storage and compute separated, object storage as the durable store, live queries for apps and agents. Because Milvus is open source, you can read the code to see how a lakebase is put together.