segments_partitions.md

Details	Why / Benefits
Partition
A way of breaking up a topic's messages into separate, distributed, "logs" Partitions can live on different nodes in the cluster If a message has no key, it is distributed "round-robin" into a partition If a message has a key, its hash is used to determine which partition it's written to	Enables scaling reads/writes across diff nodes Removes single bottlenecks Guarantees messages of the same key are always on the same partition (to guarantee order and efficiency)
Segment
A collection of messages within a partition Is the way partitions are efficiently broken down to store messages Each partition is divided into "segments" There is always an active segment to write messages into When the active segment reaches the limit (determined by your configuration), it's closed and a new active segment created	Enables a partition to store messages across many files instead of one massive file Make cleanup much more efficient