Dgraph — Distributed Native GraphQL Database

Introduction

Dgraph is a graph database that speaks GraphQL as a first-class query language. Unlike Neo4j's Cypher or Gremlin, Dgraph lets you point a GraphQL schema at the database and get a ready-to-use API — with traversals, aggregations, and filters all expressed in GraphQL itself.

With over 21,000 GitHub stars, Dgraph is designed for horizontal scale. It shards data across machines using a purpose-built Raft-based cluster (no external coordinator), making it one of the few truly distributed graph databases.

What Dgraph Does

Dgraph stores triples (subject, predicate, object) — the classic graph primitive — but exposes them as typed GraphQL objects with relationships. Schema directives like @search, @dgraph, @hasInverse, and @auth control indexing, naming, inverse edges, and row-level authorization.

Architecture Overview

Client (GraphQL / DQL / gRPC)
        |
  [Zero nodes]   <-- cluster coordinator (Raft)
        |
  [Alpha nodes]  <-- data + query workers
   sharded by predicate
   each shard replicated 3x via Raft
        |
  [Storage engine]
   Badger LSM (Go, RocksDB-like)
        |
  Horizontal scale:
   add more Alphas, Zero auto-balances

Self-Hosting & Configuration

# docker-compose.yml — 3-alpha HA cluster
version: "3.8"
services:
  zero:
    image: dgraph/dgraph:v23.1
    command: dgraph zero --my=zero:5080
    ports: ["5080:5080", "6080:6080"]

  alpha1:
    image: dgraph/dgraph:v23.1
    command: dgraph alpha --my=alpha1:7080 --zero=zero:5080
    ports: ["8080:8080", "9080:9080"]

  alpha2:
    image: dgraph/dgraph:v23.1
    command: dgraph alpha --my=alpha2:7080 --zero=zero:5080
    ports: ["8081:8080", "9081:9080"]

  alpha3:
    image: dgraph/dgraph:v23.1
    command: dgraph alpha --my=alpha3:7080 --zero=zero:5080
    ports: ["8082:8080", "9082:9080"]

# Auth example: only show posts owned by the logged-in user
type Post @auth(
  query: { rule: "query($USER: String!) { queryPost { author { name(eq: $USER) } } }" }
) {
  id: ID!
  title: String!
  author: Person!
}

Key Features

• Native GraphQL — schema-first API, no resolver boilerplate
• DQL — Dgraph Query Language for advanced traversals
• ACID transactions — distributed, Raft-replicated
• Horizontal scale — predicates sharded across Alpha nodes
• Full-text + geo indexes — search + spatial queries in-engine
• @auth rules — row-level authorization in GraphQL directives
• Cloud + self-hosted — Dgraph Cloud or Kubernetes/Docker
• Go implementation — single binary, easy ops

Comparison with Similar Tools

FAQ

Q: Dgraph vs Neo4j? A: Neo4j is the graph database incumbent with a vast Cypher ecosystem and mature tooling. Dgraph is distributed by design and GraphQL-native. Pick Neo4j for graph-heavy OLTP where Cypher is a plus; Dgraph when GraphQL and horizontal scale matter.

Q: Is Dgraph truly open source? A: Yes, Apache-2.0 for the core. Some enterprise features (binary/ACL backups, audit logging) require a paid license. Dgraph Cloud is a managed option.

Q: How does Dgraph scale? A: Data is sharded by predicate across Alpha nodes. Add more Alphas and Zero rebalances predicates automatically. Raft replication across 3 nodes gives HA per shard.

Q: Does Dgraph have a managed option? A: Yes, Dgraph Cloud (hypermodernized serverless GraphQL backend). Or self-host on Kubernetes — the Helm chart and Docker Compose examples cover most needs.

Sources

• GitHub: https://github.com/dgraph-io/dgraph
• Docs: https://dgraph.io/docs
• Company: Hypermode (acquired Dgraph)
• License: Apache-2.0