Part 1 of 8 · 1 min read
Why this page comes sixth
In this module you will
- Define capacity, goodput and headroom, and describe load and performance with the right numbers.
- Explain why percentiles beat averages and why tail latency grows with the number of calls.
- Run a load test that does not hide queueing, and read its results.
- Plan capacity for a peak with spare room for failures and releases.
- Name the common capacity myths and testing mistakes, and the safe ways to test failure.
Pages 2 to 5 said "stay under the limit". This page asks where the limit is. A limit that is guessed is usually wrong, and the guess is paid for during the worst hour of the year. The books used are "Designing Data-Intensive Applications" (DDIA), chapter 1, and "Understanding Distributed Systems" (UDS), chapters 1, 19 and 20. Where a point comes from common practice and not from these books, the text says so. The numbers are made up to show the idea, unless the text names a book.