For our Kafka Cluster (mirrored) what are partition and replication-factor specifications for each topic? We would like to maintain specs similar to Nasdaq Data Link cluster specs.
The streaming API uses 3 replicas for each topic and a 24 hour retention period on each topic.