On Efficiently Partitioning a Topic in Apache Kafka