Dragonfly Has a Free In-Memory Store — 25x Faster Than Redis With Less Memory

Redis uses 3x more memory than it needs to. Dragonfly stores the same data in 1/3 of the RAM — and handles 25x more operations.

What is Dragonfly?

Dragonfly is a modern in-memory data store compatible with Redis and Memcached APIs. Written in C++, it uses a shared-nothing architecture and novel data structures to dramatically reduce memory usage.

Why Dragonfly Is Special

1. Redis API Compatible

import redis
r = redis.Redis(host='localhost', port=6379)

# All Redis commands work
r.set('user:1', json.dumps({'name': 'Alice'}))
r.hset('stats', mapping={'views': 100, 'clicks': 42})
r.lpush('queue', 'task1', 'task2')
r.zadd('leaderboard', {'player1': 100, 'player2': 85})

2. Dramatically Less Memory

Dataset	Redis Memory	Dragonfly Memory
1M strings	85MB	30MB
1M hashes	120MB	45MB
10M keys	1.2GB	400MB

Dragonfly uses a novel hash table (Dash) that's more memory-efficient than Redis's dictionaries.

3. Multi-Threaded (Shared-Nothing)

# Uses all CPU cores automatically
docker run --ulimit memlock=-1 -p 6379:6379 docker.dragonflydb.io/dragonflydb/dragonfly

# Benchmark: 4M ops/sec on a single instance (vs ~100K for Redis)

Each thread owns its data slice. No locks, no contention.

4. Snapshotting Without Fork

Redis uses fork() to create snapshots — doubles memory during save. Dragonfly uses a novel algorithm that snapshots without forking:

# Redis: 8GB data → needs 16GB RAM during BGSAVE
# Dragonfly: 8GB data → needs ~8GB RAM during snapshot

5. Tiered Storage

dragonfly --tiered_prefix=/mnt/nvme/dragonfly

Automatically moves cold data to SSD while keeping hot data in RAM.

Performance

Benchmark	Redis 7	Dragonfly
Throughput (SET)	100K/s	4M/s
Throughput (GET)	110K/s	4.2M/s
P99 latency	0.5ms	0.3ms
Memory efficiency	Baseline	3x better
Snapshot overhead	2x memory	~0

Dragonfly vs Redis vs KeyDB

	Dragonfly	Redis	KeyDB
Architecture	Shared-nothing	Single-threaded	Multi-threaded
Memory efficiency	3x better	Baseline	Same as Redis
Snapshot	No fork needed	Fork (2x memory)	Fork
License	BSL 1.1	SSPL	BSD
Memcached compat	Yes	No	No

Getting Started

# Docker (simplest)
docker run --ulimit memlock=-1 -p 6379:6379 \
  docker.dragonflydb.io/dragonflydb/dragonfly

# Connect with redis-cli
redis-cli -p 6379
> SET hello "dragonfly"
> GET hello

The Bottom Line

Dragonfly is what Redis would be if designed today. 25x throughput, 3x memory efficiency, no-fork snapshots, and full API compatibility. If your Redis costs are growing — Dragonfly cuts them by 70%.

---

Need data solutions? I build web scraping tools. Check my Apify actors or email [e-mail].