A larger screen, mouse and keyboard make designing and running simulations much easier. Open this page on your laptop or desktop to continue.
Choose a lesson, explore its goals, then start learning.
Showing 28 of 28 lessons
Work out the chapter's Twitter QPS, size servers for the average, watch the peak break them, then size for the peak.
Change the architecture, run the simulation, then use measured results to prove each step.
The chapter's example: 300 million monthly active users, 50% of them active each day, 2 tweets per user a day. That is 150 million × 2 ÷ 24 hours ÷ 3,600 seconds ≈ 3,500 tweets a second. In this simulator one CPU core serves about 9,000 requests a second, so one server with half a core (about 4,500) covers the average. The run's peak, from 4 s to 6 s, is twice the average.
The chapter estimates peak QPS as twice the average: about 7,000 tweets a second. Keep half-core servers and add enough of them for the peak. The target is the 99.9% SLA the chapter quotes for large cloud providers.