What is the significance of num.partitions for a topic?

Partition count determines: 1) Maximum consumer parallelism (consumers in a group = partitions). 2) Write throughput (more partitions = more parallel writes to different brokers). 3) Ordering guarantee scope (ordering is only within a partition). 4) Resource overhead (each partition costs file handles, in-memory metadata, replication overhead). Guideline: target 50-100 MB/s throughput per partition. For consumer scaling: partitions ≥ peak consumer count. You can increase partitions but NOT decrease them without disrupting key-based routing.

Expected key concepts: partition count, consumer parallelism, write throughput, ordering, resource overhead, 50-100 MB/s, increase, decrease, key-based routing