Architecture lab

Challenge 6: Flash-sale catalog with a cold cacheTarget 20,000 RPS

0% busy
Shoppers
1x inst20,000 RPS
100% busy
Load balancer
4x inst20,000 RPS
83% busy
App servers
24x inst20,000 RPS
82% busy
Product cache
2x inst19,600 RPS
55% busy
Database
2x inst1,380 RPS

Design summary

Components

  • Shoppers: 1 × Users
  • Load balancer: 4 × Load balancer, 20,000 requests a second, 100% busy
  • App servers: 24 × App server, 20,000 requests a second, 83% busy
  • Product cache: 2 × Cache, 19,600 requests a second, 82% busy
  • Database: 2 × Relational database, 1,380 requests a second, 55% busy

Connections

  • Shoppers sends requests to Load balancer
  • Load balancer sends requests to App servers
  • App servers sends requests to Product cache
  • Product cache sends requests to Database
  • App servers sends requests to Database

20,000 of 20,000 requests a second served, average 41 ms, p99 98 ms, 2,040 USD a month. Challenge not yet passed: losing one instance, cold cache.