How many records will be displayed in the UI of the Data Browser. After this limit is hit, all data coming from the backend for the same request will be ignored.

How many concurrent requests are made to poll data from the broker. This is broken down by partition. If a topic has 3 partitions and we set the parallelism factor to 3 then 3 concurrent requests are issued, one per partition and the data shipped to the front end concurrently, ie, the records will show up interleaved in the UI. If the topic has only 1 partition then the parallel factor has no effect since data from the same partition is fetched in order.

This is number of records the client will try to retrieve on each poll, ie, poll batch size.

In Kinesis, there's a hard limit of 10.000 records per poll, therefore, saving this setting with a number higher than 10.000 will always return at maximum 10.000 records.