Exercise: Two Architectures, Same SLA

1. The Brief

A mobile product has an API that serves 300 million requests a month. Traffic follows a daily cycle, with a peak about four times the daily mean. Responses average 8 KB. Each request makes two reads against a managed database. The service is stateless.

The availability target is 99.9 percent, and p99 latency has to stay under 200 ms. Two designs meet both.

Architecture A, a container fleet. Containers behind a load balancer, spread across three availability zones, autoscaled between three and six instances. Each instance handles about 150 requests per second