Overview
Measures the crossover point where an in-process Go sync.Map becomes faster than a local Redis instance. Tests variable payload sizes and concurrency levels to identify the latency tradeoff.
Methodology
- Local Redis: Same host, Unix socket
- Go sync.Map: In-process with RWMutex
- Payloads: 64B, 1KB, 64KB
- Concurrency: 1–1K goroutines
Key Finding
For payloads under 1KB, sync.Map wins up to ~100 goroutines. Beyond that, Redis offers better tail latency due to connection pooling. The crossover moves with payload size.