CameoDB v0.3.7: Two-Phase Paging, Millisecond Recovery and a Windows Service
CameoDB v0.3.7 pages distributed searches in two phases for 5x faster warm queries, recovers from a crash in milliseconds and installs as a Windows service.
The engine piece by piece: storage, search, transactions and the leaderless mesh.
CameoDB fuses an ACID key-value store and a full-text engine into one binary with a single write path. These posts follow that design from the storage layer up to a multi-node mesh. The Operate page covers running it, and the query syntax shows what the search engine accepts.
CameoDB v0.3.7 pages distributed searches in two phases for 5x faster warm queries, recovers from a crash in milliseconds and installs as a Windows service.
Where CameoDB's speed comes from, and the numbers cameodb-bench measures on a single node.
Zero master nodes, no Raft, no Paxos. Consistent hashing places the data and Kademlia finds the peers.
How CameoDB keeps every write ACID across a leaderless mesh, with no consensus on the data path.
Inverted indexes, BM25 scoring, phrase queries and field boosting: how Tantivy powers search in CameoDB.
ACID key-value storage and full-text search behind one write path, with Redb as the source of truth.