Kafka streams

java library Deeply coupled to kafka Declarative - don’t need deep knowledge of network or producer consumer API

SERDE - serializer/deserrializer

• Gives you infinite stream of “strings” for instance

Aggregators transform streams into k-table (?)

• (mutable table)
• Still an infinite stream of values

Donstream consumers realize they get updates to the k-table

• This is like getting a changelog stream from a relational database
• You ask for a value as it was at a given time (defaulting to now)
• But more changes might/will come along later

A change to the keys in a stream causes a repartition. This is expensive because it involved network I/O to write the new key/value pairs to a topic and read them back out.

Streams handles divvying up work (paritions) into tasks and spreads them across threads.