CSE Roadmap

Core tier · about 3 hours · 8 parts

Part 5 of 8 · 1 min read

Unexpected load

Every system has a wall, and spikes find it.

Every system can handle only so much load without scaling. A slow rise gives time to add machines. A sudden spike does not. UDS lists reasons the rate or the kind of requests can change fast:

  • Seasonality. Different time zones hit the service at different hours.
  • Expensive requests. Some requests cost far more than others, and scrapers may pull data at a speed no person could.
  • Malicious traffic. A distributed denial of service (DDoS) attack tries to fill the service's bandwidth so real users cannot get in.

The book says that to survive unexpected load, preparation has to come before the event. It also notes that the simple techniques of page 4 let systems it ran double their traffic almost overnight during the 2020 pandemic without incident.

Check yourself: 1. Name two reasons load can rise suddenly. 2. Why is scaling out in the middle of a spike often too slow? 3. What does a scraper do that an ordinary user does not?