NebulaGraph

What it is

NebulaGraph is an open-source, distributed graph database built in C++ for storing and querying large-scale graphs — think billions of vertices and edges with millisecond-level query latency. It uses a shared-nothing architecture separating storage (RocksDB-backed) from compute/query layers, and speaks its own query language, nGQL (openCypher-compatible). It's positioned in CNCF's landscape under App Definition and Development, though CNCF maturity level is not publicly available, suggesting it may not be a CNCF-hosted project or the data hasn't been indexed.

Who builds it and why

The project originated from vesoft inc., a company founded by ex-Facebook engineers who worked on graph infrastructure at that scale. That heritage shows in the design goals: horizontal scalability and separation of storage/compute, patterns borrowed directly from hyperscaler internal systems. 119 contributors on GitHub indicates a real community beyond the founding company, though the concentration of commit authorship (not visible from stars/contributor count alone) would need direct repo inspection to assess how much is vesoft-driven vs. community-driven. Founding date is not publicly available in the data provided, but the project has been active in the graph database space for several years.

Production readiness signal

12,253 GitHub stars and a commit as recent as 2026-05-18 show the project is actively maintained, not abandoned. Latest release v3.8.0 suggests a mature versioning cadence — this isn't a 0.x experimental tool. Apache-2.0 licensing removes legal friction for commercial use. That said, "production readiness" for a distributed graph database is a heavier claim than for most infra tools: you need to evaluate operational maturity (backup/restore tooling, multi-region replication, upgrade paths between major versions) directly against your workload, not just star count. CNCF maturity level being unavailable means you can't lean on that signal — you're on your own for vetting.

Who should use this

• Teams with genuinely large graph datasets (billions of edges) where a relational database's JOIN performance collapses — fraud detection networks, knowledge graphs, social/recommendation graphs, supply chain traceability.
• Engineering orgs comfortable operating a distributed C++ system in production — this isn't a managed SaaS-first product, so you need in-house expertise in distributed systems debugging.
• Organizations that specifically need horizontal scale-out beyond what a single-node graph DB (like Neo4j Community) can offer, and are willing to trade some ecosystem maturity for that scale.

Who should NOT use this

• Teams with small-to-medium graphs (under tens of millions of edges) — Neo4j or even PostgreSQL with recursive CTEs will be simpler to operate and hire for.
• Anyone wanting a managed cloud offering with minimal ops burden — verify current managed service availability directly, as this data doesn't confirm one exists at the maturity level of Neo4j Aura or AWS Neptune.
• Teams without distributed systems operational experience — running a shared-nothing C++ storage/compute cluster is not a weekend project, and documentation/tooling maturity for edge cases (split-brain, compaction tuning) is inconsistent across graph databases in general.
• Projects needing strict ACID multi-statement transactions as a primary requirement — verify current transaction guarantees against your consistency needs before committing.

Alternatives

• Neo4j — the incumbent graph database with the largest ecosystem, better tooling and documentation, but weaker horizontal scale-out story in its open-source edition.
• Amazon Neptune — fully managed graph database (property graph + RDF) if you want zero ops burden and are already on AWS, at the cost of vendor lock-in and less control.
• JanusGraph — another distributed, open-source graph database (Apache-2.0) built on pluggable storage backends (Cassandra, HBase, Bigtable), a closer architectural peer to NebulaGraph than Neo4j is.

Pricing

Fully open source under Apache-2.0 — no licensing cost for self-hosting. vesoft inc. may offer a commercial cloud/enterprise product, but pricing and feature-gating details for any such offering are not publicly available in the data provided; verify directly with vesoft before assuming feature parity between OSS and any paid tier.