# KeyDB — Multithreaded Drop-In Redis Replacement

> A fully Redis-compatible fork that is multithreaded by design, supports active-active replication, and delivers up to 5x the throughput of Redis on modern multi-core servers without changing a line of client code.

## Install

Save as a script file and run:

# KeyDB — Multithreaded Drop-In Redis Replacement

## Quick Use
```bash
# Run multithreaded KeyDB with 4 worker threads
docker run --rm -p 6379:6379 eqalpha/keydb 
  keydb-server --server-threads 4

# Any Redis client works unchanged
redis-cli PING
redis-cli SET hello world
redis-cli GET hello

# Active replication between two nodes
keydb-server --port 6379 --active-replica yes --replicaof 10.0.0.2 6379
keydb-server --port 6379 --active-replica yes --replicaof 10.0.0.1 6379
```

## Introduction
KeyDB is a high-performance fork of Redis that adds native multithreading, active-active replication and FLASH tiered storage while staying 100% compatible with the Redis protocol, data types and client libraries. It is for teams that love Redis but have outgrown its single-threaded event loop on modern 32-core servers.

## What KeyDB Does
- Runs the Redis protocol on multiple worker threads that share the keyspace.
- Supports active-active master-master replication with last-writer-wins conflict resolution.
- Adds FLASH, an optional NVMe-backed tier that keeps warm keys in RAM and cold keys on SSD.
- Exposes all Redis data structures: strings, hashes, lists, sets, sorted sets, streams, hyperloglogs, bitmaps, geo.
- Ships Lua scripting, pub/sub, cluster mode, ACLs and TLS — exactly like Redis.

## Architecture Overview
Each worker thread handles its own connections and runs an event loop. Shared data structures use fine-grained spinlocks and a global key lock that is held only briefly, letting many cores execute commands concurrently. Active replication exchanges writes both ways continuously, so two masters in different regions stay in sync without a manual failover. FLASH uses an LSM-backed layer so cold entries spill to disk transparently.

## Self-Hosting & Configuration
- Install from apt (`eqalpha/keydb` PPA), yum, Docker image, or build from source.
- Set `server-threads` to ~N physical cores minus 1 for the best throughput.
- Enable `multi-master yes` + `active-replica yes` for read/write on both regions.
- Use `storage-provider flash /data/keydb-flash` to unlock FLASH tiered storage.
- Tune `maxmemory-policy`, `tcp-backlog` and `tcp-keepalive` as you would with Redis.

## Key Features
- Up to 5x Redis throughput on 16+ core machines with no client changes.
- Multi-master replication with sub-millisecond convergence in a single DC.
- FLASH tiered storage for caches far larger than RAM.
- Redis 7-level command compatibility, Sentinel and Cluster protocols supported.
- Drop-in: existing monitoring (redis_exporter, redis-cli) works unchanged.

## Comparison with Similar Tools
- **Redis 7** — canonical upstream; now I/O-threaded but not yet fully multithreaded.
- **Valkey** — Linux Foundation fork of Redis 7.2; different governance, single-threaded core.
- **DragonflyDB** — from scratch multi-threaded Redis clone with shard-per-thread design.
- **Skytable / KeyDB Pro** — alternative takes; smaller ecosystems.
- **Memcached** — simpler KV cache, no data structures or persistence.

## FAQ
**Q:** Is KeyDB really a drop-in replacement?
A: Yes — Redis 6/7-level protocol and RDB/AOF formats are compatible. Swap the binary and keep your clients.
**Q:** Do I still need Redis Cluster?
A: For sharding across many nodes, yes — KeyDB supports the same cluster protocol.
**Q:** Active replication: how are conflicts resolved?
A: Last-writer-wins per key, with timestamps on each write. Ideal for session/cache workloads.
**Q:** Is KeyDB maintained?
A: Yes — Snap bought KeyDB in 2022 and continues to release patches on GitHub.

## Sources
- https://github.com/Snapchat/KeyDB
- https://docs.keydb.dev/

---
Source: https://tokrepo.com/en/workflows/afb23d9f-3920-11f1-9bc6-00163e2b0d79
Author: Script Depot