OrbitDB — Peer-to-Peer Serverless Database on IPFS

Introduction

OrbitDB is a peer-to-peer database that uses IPFS for data storage and libp2p for networking. It requires no central server — data replicates automatically between peers using CRDTs to resolve conflicts, making it ideal for decentralized and offline-first applications.

What OrbitDB Does

• Provides multiple database types: key-value, documents, events log, and feed
• Replicates data peer-to-peer over IPFS without central coordination
• Resolves conflicts automatically using CRDTs (no manual merge needed)
• Works offline and syncs when connectivity is restored
• Supports access control via cryptographic identities

Architecture Overview

OrbitDB stores data as an append-only operation log on IPFS. Each database is a CRDT that merges concurrent writes deterministically. The log entries reference each other via content-addressed IPFS hashes forming a DAG. Libp2p handles peer discovery and pub/sub messaging to notify peers of new entries. Each peer maintains a local copy and can reconstruct the full state from the operation log.

Self-Hosting & Configuration

• Install via npm alongside a Helia (IPFS) node instance
• No server required — each application instance is a peer
• Configure IPFS swarm addresses for peer discovery on custom networks
• Set access controller to restrict who can write (default: creator only)
• Persist data locally using the default LevelDB-backed blockstore

Key Features

• Fully decentralized with no single point of failure
• Offline-first: works without internet, syncs when peers reconnect
• Multiple database models cover common use cases (docs, KV, log)
• Content-addressed storage ensures data integrity via IPFS
• Identity-based access control using public key cryptography

Comparison with Similar Tools

• GunDB — Also P2P but uses its own protocol; OrbitDB leverages the IPFS ecosystem
• PouchDB/CouchDB — Sync-capable but relies on a central CouchDB server for replication
• Ceramic — Decentralized data streams focused on identity; OrbitDB is general-purpose
• Automerge — CRDT library without built-in networking; OrbitDB bundles storage and sync

FAQ

Q: Do I need to run an IPFS node? A: Yes, OrbitDB requires a Helia (JS-IPFS) instance, but it runs in-process with your app.

Q: How does access control work? A: Each database has an access controller that checks the writer's cryptographic identity before accepting writes.

Q: Can OrbitDB scale to many peers? A: It works well for dozens to hundreds of peers. Very large networks may need topic sharding for pub/sub efficiency.

Q: Is data encrypted? A: Data on IPFS is content-addressed but not encrypted by default. You can encrypt payloads at the application layer before writing.

Sources

• https://github.com/orbitdb/orbitdb
• https://orbitdb.org