Bedrock and UnisonDB are racing to be AI-ready data stores. Bedrock is a modular, WAN-replicated database based on SQLite [1], while UnisonDB unifies storage and streaming into a single log-native core [2].
Replication semantics Bedrock relies on WAN replication across nodes, emphasizing modularity to span distributed setups [1]. In contrast, UnisonDB uses a WAL-based replication model, with WAL streams delivered via gRPC and followers tailing in real time across 100+ nodes [2].
Multi-model support UnisonDB advertises multi-model support—KV, wide-column, and large objects (LOB) in one atomic transaction [2]. Bedrock is described as modular and SQLite-based, but multi-model details aren’t spelled out in the available material [1].
Deployment & tradeoffs Bedrock’s WAN-replication, built on SQLite, points to distributed deployments that blend familiar SQLite foundations with cross-site replication [1]. UnisonDB is edge-friendly too—replicas can go offline and resync, and writes emit real-time notifications via ZeroMQ as part of its reactive design [2].
Performance note A key tradeoff: write throughput in UnisonDB is lower than pure LSM stores because global ordering is required for replication safety [2].
Closing thought: if you want real-time, edge-spanning replication with a unified log core, UnisonDB fits. If you prefer a modular, SQLite-based path for WAN deployments, Bedrock is the pick.
References
Bedrock, a modular, WAN-replicated database based on SQLite
Bedrock is a modular, WAN-replicated database based on SQLite, emphasizing modular design and wide-area data replication.
View sourceUnisonDB blends storage and streaming in a log-based core, avoiding CDC/Kafka; WAL replication, B+Tree storage, multi-model KV.
View source