Which join scenario allows joining a topic with 5 partitions to a topic with 3 partitions without co-partitioning?

Which join scenario allows joining a topic with 5 partitions to a topic with 3 partitions without co-partitioning?

• A. KStream-to-KStream join
• B. KTable-to-KTable join
• C. KStream-to-KTable join
• D. KStream-to-GlobalKTable join

Answer: (D)

In Kafka Streams, how a join is performed depends on partitioning. For most joins, you need co-partitioning: the two sides must be partitioned so that messages with the same join key land on the same partition across both topics, which usually means matching the number of partitions or repartitioning one side. The twist is joining a KStream to a GlobalKTable. A GlobalKTable is replicated to every application instance, so each instance has a full copy of the table and can look up keys locally without needing the input stream to be partitioned to match the table. Because of this replication, you can join a topic with 5 partitions to a GlobalKTable derived from a topic with 3 partitions without performing any co-partitioning or repartitioning. That’s why the best match is the KStream-to-GlobalKTable join. The other join types rely on co-partitioning between their respective topics, so differing partition counts would require repartitioning to align keys. Keep in mind that GlobalKTable introduces data replication across all instances, which has its own trade-offs in memory usage and update cost.

In Kafka Streams, how a join is performed depends on partitioning. For most joins, you need co-partitioning: the two sides must be partitioned so that messages with the same join key land on the same partition across both topics, which usually means matching the number of partitions or repartitioning one side.

The twist is joining a KStream to a GlobalKTable. A GlobalKTable is replicated to every application instance, so each instance has a full copy of the table and can look up keys locally without needing the input stream to be partitioned to match the table. Because of this replication, you can join a topic with 5 partitions to a GlobalKTable derived from a topic with 3 partitions without performing any co-partitioning or repartitioning.

That’s why the best match is the KStream-to-GlobalKTable join. The other join types rely on co-partitioning between their respective topics, so differing partition counts would require repartitioning to align keys. Keep in mind that GlobalKTable introduces data replication across all instances, which has its own trade-offs in memory usage and update cost.