Extremely fast Redis-compatible cache

High Performance Key Value Database, up to 4.5× faster than Redis.

Redistill is a high-performance, in-memory key-value database that speaks the Redis protocol and delivers record-setting single-instance throughput with sub-millisecond latency.

Get started in 5 minutes View on GitHub
Throughput
9.07M ops/s
Up to 4.5× Redis, 1.7× Dragonfly
Latency (p50)
0.48 ms
5× lower than Redis
Cost efficiency
50–83% less
Fewer instances for same load

Why Redistill

Performance-first, drop-in compatible with Redis clients.

Designed as a distilled version of Redis focused on high-performance key-value workloads, Redistill eliminates persistence overhead while keeping the parts you rely on every day.

High-performance KV

Fastest single-instance KV database

Multi-threaded architecture, lock-free reads with DashMap, zero-copy buffers, and batched I/O combine to deliver over 9M operations per second on a single instance.

  • Multi-core utilization via Tokio runtime
  • 2048–4096 shards for reduced contention
  • Optimized for pipelined workloads

Redis protocol

Works with existing Redis clients

Implements the Redis RESP protocol so you can point redis-cli, redis-py, ioredis, go-redis and more at Redistill without changing your application code.

  • Core commands: SET, GET, MGET, HSET, TTL, SCAN
  • INFO, AUTH and health monitoring
  • Optional snapshot persistence for warm restarts

Cost efficiency

Fewer instances, lower infra cost

With up to 4.5× Redis throughput per instance, you can support the same load with significantly fewer servers, cutting infrastructure costs by 50–83% in typical scenarios.

  • Great fit for read-heavy, cacheable data
  • Elastic scaling via client-side sharding
  • Configuration tuned for modern hardware

Use cases

Built for high-throughput, low-latency workloads.

Redistill shines as a high-speed cache and session store where data can be regenerated and latency directly impacts user experience.

Session storage

Store and read millions of session keys per second with automatic TTL expiration and predictable sub-millisecond latency.

  • 1M+ ops/sec for auth/session tokens
  • Simple key patterns like session:<id>
  • Easy integration with popular web frameworks

API response caching

Offload hot database queries to an in-memory cache with 50–150× faster response times for repeated requests.

  • JSON/HTML responses in the 1–50KB range
  • Typical hit rates above 95%
  • Perfect for product catalogs and user profiles

Rate limiting

Use atomic counters with TTL to implement per-user, per-API or per-IP rate limits with minimal extra latency on hot paths.

  • INCR/EXPIRE-based limits
  • Millions of counters with tiny memory footprint
  • Perfect fit for API gateways and edge proxies

Real-time leaderboards & counters

Track impressions, scores, and lightweight metrics where occasional data loss is acceptable but performance is critical.

  • Increments and hash fields for aggregates
  • TTL-based cleanup of stale keys
  • Rebuildable from primary databases

Features

Production-ready, focused feature set.

A lean, well-tuned subset of Redis designed for the vast majority of cache workloads, with security and observability built in.

Core data model

High-performance string and hash storage with TTLs, counters, and non-blocking key scanning.

SET / GET / MGET / MSET INCR / DECR / INCRBY HSET / HGET / HGETALL TTL / PTTL / PERSIST SCAN / DBSIZE

Security & reliability

Password authentication, TLS termination, connection limits, and health checks so you can run Redistill safely in production.

AUTH TLS/SSL Max connections Connection rate limiting Graceful shutdown

Observation & tuning

Rich configuration and metrics so you can dial in performance for your workload and monitor behavior over time.

INFO metrics HTTP health endpoint Eviction policies (LRU, random, S3-FIFO) Snapshot persistence (optional)

Get started

Install Redistill and run your first benchmark.

Use Docker, Homebrew, direct binaries, or build from source. Redistill speaks the Redis protocol, so your existing Redis tools and clients just work.

Install & run

The fastest way to get going is Docker. You can also install via Homebrew on macOS or download platform-specific binaries. 

docker run -d --name redistill \
  -p 6379:6379 shahidontech/redistill:latest

redis-cli ping
# PONG

See the Quick Start for Homebrew, Linux, Windows, and source builds.

Benchmark

Validate performance on your own hardware using redis-benchmark or memtier_benchmark. 

redis-benchmark -t set,get \
  -n 1000000 -c 50 -P 16 -q

For complete competitive results vs Redis and Dragonfly, visit the Benchmarks page.