The core performance and reliability metrics.
Latency vs throughput
Latency is the time one request takes. Throughput is how many requests you finish per unit time. Optimising one can hurt the other (batching raises throughput but adds latency). Track percentiles (p50, p95, p99), not averages.
Availability and the nines
Availability = uptime / total time.
• 99% ≈ 3.65 days down per year
• 99.9% ≈ 8.8 hours per year
• 99.99% ≈ 52 minutes per year
• 99.999% ≈ 5 minutes per year
Each extra nine is much more expensive.
Achieving availability
Remove single points of failure with redundancy: multiple servers, multiple zones, database replicas, health checks and automatic failover.