Nauka is a distributed storage engine designed to keep files available when machines, disks, or even entire regions fail. It runs as a single binary and can form a cluster with two commands: the first machine starts the cluster, and additional nodes can be provisioned over SSH. Nauka divides files into Reed-Solomon shards and distributes them across cluster nodes.

Its default 4+2 configuration can recover data if any two of six shards are lost, with 50% storage overhead. BLAKE3 checks detect corrupted shards, while a continuous scrubber regenerates missing data. Nauka also refills returning nodes and drains nodes being removed.

For geographically dispersed clusters, it measures network latency to place shards across locations and supports monthly egress budgets for metered links. The engine exposes an HTTP API, with an S3-compatible endpoint available through an opt-in build flag. Nauka is deliberately focused on storage rather than user-facing applications; products and services can be built on top of its engine.

Its documentation covers quickstarts, durability and consistency, deployment, and CLI commands