NoKV provides an open-source namespace metadata substrate designed for distributed filesystems, object storage, and AI dataset workloads. It implements a "stateless schema layer + transactional KV" pattern, offering native metadata primitives as first-class server-side operations. This approach benefits developers building DFS frontends, S3 gateways, or AI dataset services by delivering significantly higher performance and a unified source of truth for namespace metadata, abstracting away the complexities of underlying storage.

How It Works

NoKV features a vertically integrated architecture, encompassing its own LSM-tree storage engine (with an ART memtable), Raft consensus, Percolator MVCC, and a coordinator. This deep integration allows for the implementation of namespace-specific operations like ReadDirPlus, WatchSubtree, SnapshotSubtree, and RenameSubtree as native server-side primitives, rather than client-side compositions over generic Get/Put/Scan calls. This design yields substantial performance advantages and ensures a single, replicated event log for critical metadata states such as mount lifecycle, subtree authority, snapshot epochs, and quota fences.

Quick Start & Requirements

• Primary install/run: Docker Compose (docker compose up -d), local cluster script (./scripts/dev/cluster.sh), or embedded Go API.
• Prerequisites: Docker, Go (for building/embedded usage).
• Links:
  • Getting Started: docs/getting_started.md
  • Full Documentation: docs/SUMMARY.md
  • Demo: scripts/demo/redis-smoke.sh

Highlighted Details

• ReadDirPlus operations are up to 42.5x faster compared to generic KV calls.
• WatchSubtree provides sub-second end-to-end change feed latency for prefix-scoped metadata watches.
• The underlying KV layer demonstrates superior performance over Badger and Pebble in YCSB benchmarks.
• Key metadata primitives like ReadDirPlus, WatchSubtree, SnapshotSubtree, and RenameSubtree are first-class server-side operations.

Maintenance & Community

The project appears actively developed, with detailed documentation and recent benchmarks. No specific community channels (like Discord or Slack) or notable external contributors/sponsorships are listed in the provided README.

Licensing & Compatibility

NoKV is licensed under the Apache-2.0 license. This license is permissive and generally compatible with commercial use and linking within closed-source projects.

Limitations & Caveats

NoKV is explicitly positioned as a metadata "substrate," not a complete filesystem server; it does not ship with FUSE drivers or S3 gateways out-of-the-box. Its internal KV layer is optimized for its metadata service and is not intended as a direct replacement for general-purpose distributed KV stores like TiKV or FoundationDB.