Part 1 of 7 · 1 min read
Why this page comes third
In this module you will
- Explain why a log-based broker keeps messages after they are read, and what that makes possible.
- Describe a topic, a partition and an offset, and say what order is guaranteed and what is not.
- Explain how a consumer group splits a topic and why the number of partitions caps the parallelism.
- Choose a partition key so that related messages stay in order.
- Work out how long a log can buffer messages and how to notice a slow consumer.
Page 1 described queues that delete messages once they are handled. That model cannot replay history, cannot add a new reader that sees the past, and loses order when work is shared. A log keeps the messages and fixes all three. Most large streaming systems are built on it, and the next four pages assume it. The books used are "Designing Data-Intensive Applications" (DDIA) and "Understanding Distributed Systems" (UDS). The numbers are made up to show the idea, unless the text names a book.