Kafka diskalakan untuk throughput massive (jutaan events/detik) melalui partitioning, menambah brokers dan consumers, dan tuning. Memahami cara Kafka menskalakan penting untuk deployment dengan volume tinggi.
Scaling levers
✓ PARTITIONS → the primary scaling unit: more partitions → more parallelism (producer and
consumer) → distribute data and load across brokers and consumers
✓ BROKERS → add brokers to the cluster → more storage, throughput, and capacity (spread
partitions across more machines)
✓ CONSUMERS → add consumers to a group (up to partition count) → parallel consumption
✓ PRODUCERS → batching, compression, multiple producers → high write throughput
→ Kafka scales HORIZONTALLY (add partitions/brokers/consumers)
