Kafka Overview（2）- partition

Note

For the images to display correctly, switch to a white reading background.

Related

• Kafka Overview（1）- Architecture and Topics

• Kafka Overview(3) - Consumer Group

• Kafka Overview(4) - Commit Log & offset

• Kafka Overview(5) - Process of sending

Partitions

Within the Kafka cluster, topics are divided into partitions , and the partitions are replicated across brokers. From each partition, multiple consumers can read from a topic in parallel. It's also possible to have producers add a key to a message-all messages with the same key will go to the same partition.

While messages are added and stored within partitions in sequence,messages without keys are written to partitions in a round robin fashion.By leveraging keys,you can guarantee the order of processing for messages in Kafka that share the same key. This is a particularly useful feature for applications that require total control over records. There is no limit on the number of Kafka partitions that can be created(subject to the processing capacity of a cluster).

Topic - Partitions

Kafka write scalability—showing concurrent replication to followers

Kafka read scalability—partitions enable concurrent consumers