Jay Kreps
Jay Kreps
Jul 4, 2017 · 1 min read
  1. Kafka is distributed! In CAP terms Kafka is CP: it is always consistent but will not accept writes from the minority part of a partitioned network.
  2. Not so, Kafka support a logical notion of parallelism, the partition. The consumption of each partition is single threaded but there are lots of them. Kafka handles the assignment of these to the active consumers. This means you don’t need many consumers competing over one partition as you describe.

    Jay Kreps

    Written by

    Jay Kreps

    Co-founder of @confluentinc (http://confluent.io). Co-creator of @apachekafka. Taller than I look on Twitter.