CongraphDB Performance Benchmark v0.2.1

A comprehensive benchmark suite comparing CongraphDB v0.2.1 against industry-standard graph and vector databases.

📊 Live Results • Quick Start • Documentation • v0.2.1 Features

---

📊 Quick Summary

Benchmark Metrics

Metric	Description	What It Measures
⚡ Ingestion Rate	Nodes/edges per second	Write throughput & data loading performance
🔍 Query Latency	k-hop traversal (1-5 hops)	Graph navigation speed at varying depths
🧮 Algorithmic Throughput	PageRank execution time	Graph algorithm performance
💾 Memory Footprint	Peak RSS during operations	Memory efficiency & resource usage
🚀 Cold vs. Warm Start	First query vs. cached	Startup overhead & caching effectiveness

Test Scales

Scale	Nodes	Edges	Use Case
🔹 Small	10K	50K	Quick testing & development
🔸 Medium	100K	1M	Standard production workload
🔺 Large	1M	10M	Enterprise-scale scenarios

---

v0.2.1 Features

New in v0.2.1: Extended benchmarks for CongraphDB's hybrid engine, including parallel HNSW ingestion, batch vector search, and Graph RAG algorithms:

Benchmark	Description	Metrics
🤖 Vector Search	Parallel HNSW ingestion & batch similarity search	inserts/sec, search_latency_ms, recall
🧠 Graph RAG	Fact/Entity retrieval & Graph algorithms	ops/sec, traversal_depth, memory_rss
🔒 Transactions	WAL durability & OCC concurrency overhead	latency_ms, throughput_tps, recovery_time
🗄️ Query Engine	UNWIND & Parameterized query performance	queries/sec, planning_time_ms
📊 Algorithms	PageRank, Leiden, Closeness, Louvain	algorithm_time_ms, scaling_factor

Algorithm Benchmarks

Category	Algorithms
Centrality	PageRank, Betweenness, Closeness, Degree
Community Detection	Louvain, Leiden, Spectral, SLPA, Infomap, Label Propagation
Traversal	BFS, DFS, Dijkstra
Analytics	Triangle Count, Connected Components, SCC

Running v0.1.13 Benchmarks

# Run all v0.1.13 benchmarks
npm run benchmark:v0113

# Run specific v0.1.13 benchmark
npm run benchmark:dml      # DML operations
npm run benchmark:api      # API comparison
npm run benchmark:vector   # Vector search (v0.1.x path)

# Run v0.2.1 Graph RAG benchmarks
npm run benchmark:v021     # All v0.2.1 benchmarks
npm run benchmark:rag      # Graph RAG specifically

New CLI Options (v0.2.1)

Option	Description	Default
--v021	Enable v0.2.1 benchmarks	false
--v021-benchmarks <list>	Comma-separated v0.2.1 benchmarks: graph-rag	all

Legacy CLI Options

Option	Description	Default
--v0113	Enable v0.1.13 benchmarks	false
--v0113-benchmarks <list>	Comma-separated v0.1.13 benchmarks	all

---

Overview

This benchmark validates CongraphDB's performance advantages as an Embedded Graph Database for Node.js applications.

🏆 Competitors

Database	Type	Language	Architecture
CongraphDB	🆚 Target	TypeScript	Embedded Graph DB
Neo4j	Industry Standard	Java	Client-Server Graph DB
Kùzu	Embedded Graph	C++	Embedded Graph DB
Graphology	In-Memory	JavaScript	Pure-JS Graph Library
LevelGraph	RDF Store	JavaScript	JS RDF Triple Store
Better-SQLite3	Relational	C	Embedded SQL Database

🚀 Quick Start

# Install dependencies
npm install

# Run medium-scale benchmark with default engines
npm run benchmark

# Run specific scale
npm run benchmark:small   # 10k nodes / 50k edges
npm run benchmark:medium  # 100k nodes / 1M edges
npm run benchmark:large   # 1M nodes / 10M edges

# Run specific engines
node dist/cli.js run --engines congraph,neo4j,sqlite

# View help
node dist/cli.js run --help

Environment Setup

Neo4j (Optional)

To benchmark against Neo4j, you need a running Neo4j instance:

# Using Docker
docker run -d \
  --name neo4j \
  -p 7687:7687 -p 7474:7474 \
  -e NEO4J_AUTH=neo4j/your_password \
  neo4j:5.26-community

Then create a .env file:

NEO4J_URI=bolt://localhost:7687
NEO4J_USER=neo4j
NEO4J_PASSWORD=your_password

Kùzu (Optional)

To benchmark against Kùzu, install the Node.js bindings:

# Follow https://github.com/kuzudb/kuzu for installation instructions
npm install kuzu

📈 Results

Interactive Documentation

View interactive benchmark results with charts and visualizations

✨ Features:

• 📊 Interactive charts for all metrics
• 🏆 Performance leaderboard
• ⚖️ Engine comparisons
• 📅 Historical trends

🖥️ Local Documentation

# Serve documentation locally
npm run docs:serve

# Build documentation
npm run docs:build

# Deploy to GitHub Pages
npm run docs:deploy

Raw Results

Benchmark results are also saved to the results/ directory:

• benchmark-*.json - Full results in JSON format
• benchmark-*.csv - Summary results in CSV format

🏅 Leaderboard

Results from the latest benchmark run

💡 Tip: Run npm run benchmark to generate fresh results!

Engine	Scale	⚡ Ingestion	🔍 1-hop	🔍 3-hop	🔍 5-hop	🧮 PageRank	💾 Memory
CongraphDB	Small	45,200/s	0.4ms	2.1ms	8.4ms	120ms	115MB
Neo4j	Small	12,800/s	1.2ms	5.4ms	24.1ms	450ms	1.2GB
SQLite	Small	8,500/s	0.8ms	8.2ms	45.3ms	1.8s	45MB
Kùzu	Small	38,400/s	0.5ms	2.5ms	10.2ms	180ms	180MB

🤖 Vector & Graph RAG Performance (v0.2.1)

Results for 100,000 vectors (1536-dim) on a 16-core workstation:

Operation	Complexity	CongraphDB (v0.2.1)	Improvement
HNSW Ingestion	Parallel	14,850 nodes/s	5.2x vs v0.1.x
Similarity Search	k=10, ef=50	1.2ms	12x vs Cypher-only
Batch Search	100 queries	48ms	8.5x vs Sequential
Graph RAG Search	Dual-Level	12.4ms	New Feature
Context Retrieval	3-hop Paths	8.2ms	Optimized Pathing

Architecture

congraphdb-benchmark/
├── bin/                 # CLI entry point
├── src/
│   ├── engines/         # Database adapters
│   │   ├── congraph.ts  # CongraphDB adapter
│   │   ├── neo4j.ts     # Neo4j adapter
│   │   ├── kuzu.ts      # Kùzu adapter
│   │   ├── sqlite.ts    # Better-SQLite3 adapter
│   │   ├── graphology.ts # Graphology adapter
│   │   └── levelgraph.ts # LevelGraph adapter
│   ├── benchmarks/      # v0.1.10 specialized benchmarks
│   │   ├── api-comparison.ts  # Cypher vs JavaScript API
│   │   ├── dml-operations.ts  # DML operations tests
│   │   ├── persistence.ts      # Storage mode tests
│   │   ├── statistics.ts       # Query statistics overhead
│   │   ├── vector-search.ts    # HNSW vector search benchmarks (NEW)
│   │   └── optimizer.ts        # Query optimizer benchmarks (NEW)
│   ├── core/
│   │   ├── suite.ts     # Test suite executor
│   │   └── recorder.ts  # Metrics & export
│   └── data/
│       └── academic.ts  # Synthetic data generator
├── docs/                # Interactive documentation
│   ├── index.md         # Landing page
│   ├── results/         # Results pages with charts
│   ├── engines/         # Engine comparisons
│   └── methodology/     # Testing methodology
├── data/                # Benchmark data for docs
│   ├── latest.json      # Latest results
│   └── history.json     # Historical trends
├── results/             # Benchmark output
├── scripts/             # Documentation generation
├── overrides/           # Custom styles/charts
└── .github/workflows/   # CI automation

Data Generation

The benchmark uses a Synthetic Academic Citation Network resembling the Cora/PubMed datasets:

• Nodes: Papers with metadata (title, category, year, authors, venue)
• Edges: Citation relationships with preferential attachment
• Categories: AI, ML, DB, IR, CV, NLP

Contributing

When adding new engines or benchmarks:

1. Implement the EngineAdapter interface in src/engines/
2. Add the engine to the CLI parser in src/cli.ts
3. Update this README with your results

License

MIT

---

Last updated: 2026-04-17 (v0.1.13)