Cachee Benchmark Suite

Open-source cache performance benchmark framework for comparing Cachee.ai, Redis, Memcached, ElastiCache, and other caching solutions.

Latest Results

Metric	Cachee	Redis	Memcached	ElastiCache
GET Latency	17 ns	1.0 ms	0.8 ms	1.2 ms
SET Latency	22 ns	1.1 ms	0.9 ms	1.3 ms
P99 Latency	24 ns	2.5 ms	1.8 ms	3.0 ms
Throughput	59M ops/s	~250K ops/s	~300K ops/s	~200K ops/s
Hit Rate	98.1%	~95%	~93%	~95%

Full methodology and reproduction steps: cachee.ai/benchmark-methodology

Quick Start

# Clone and run benchmarks
git clone https://github.com/HapPhi/cachee-benchmarks.git
cd cachee-benchmarks
docker-compose up -d
./run-benchmarks.sh

Benchmark Categories

Latency Benchmarks

• Single-key GET/SET — Median, P95, P99, P999
• Batch operations — MGET/MSET with 10, 100, 1000 keys
• Hot-key contention — Zipfian distribution, 80/20 access pattern

Throughput Benchmarks

• Sustained load — 1, 10, 100, 1000 concurrent connections
• Burst capacity — 10x traffic spikes sustained for 60 seconds
• Mixed workload — 80% read / 20% write realistic traffic

Trading-Specific Benchmarks

• Order lifecycle — 14-lookup sequence simulating pre-trade risk checks
• Market data fan-out — 1:N subscriber reads from shared state
• Smart order routing — Venue selection with real-time latency tables

See trading solutions for how Cachee achieves 17ns reads in production trading environments.

AI/ML Benchmarks

• KV cache lookup — LLM inference serving with shared prefix optimization
• Embedding retrieval — RAG pipeline feature store access patterns
• Agent chain — Multi-model orchestration with 10-50 cache reads per request

See AI solutions for how Cachee eliminates the GPU memory wall.

Supported Backends

Backend	Config Key	Notes
Cachee.ai	cachee	L1 in-process + L2 Redis
Redis	redis	Standalone or Cluster
Memcached	memcached	Single node
AWS ElastiCache	elasticache	Redis-compatible
Azure Cache	azure	Redis-compatible
GCP Memorystore	gcp	Redis-compatible
Upstash	upstash	Serverless Redis

See all supported integrations

Configuration

# benchmark-config.yaml
backends:
  cachee:
    host: cachee-proxy
    port: 6380
  redis:
    host: redis
    port: 6379

scenarios:
  - name: single-key-get
    operations: 1000000
    concurrency: 100
    key_distribution: zipfian

Results Output

Results are written to results/ as JSON and rendered as Markdown tables. See full benchmark results for production numbers.

Contributing

1. Fork the repo
2. Add your backend adapter in backends/
3. Run ./run-benchmarks.sh --backend=yours
4. Open a PR with results

Links

• Website: cachee.ai
• Documentation: cachee.ai/docs
• Benchmarks: cachee.ai/benchmarks
• How It Works: cachee.ai/how-it-works
• Pricing: cachee.ai/pricing
• Blog: cachee.ai/blog

License

Apache 2.0