Observability & Resilience

Why it matters

Real systems fail and slow down. You need metrics to see it and controls to keep the system alive when it happens.

SLOs and tail latency

• SLOs define what "good" looks like
• Tail latency (p99/p999) drives user experience
• Error budgets tell you how fast you're burning reliability

Resilience controls

• Retries can amplify load
• Circuit breakers stop cascades
• Backpressure & load shedding protect the core

What you will build

• P99 + burn-rate metrics
• Circuit breaker simulation
• Backpressure queue simulation
• Token bucket rate limiter
• Retry backoff scheduling