While it is true that MapR Streams doesn’t limit the number of topics in a stream and MapR Streams provides for massive scale, reality must prevail. For example, if a single stream contains millions of topics, consumers that attempt to read from these topics will generate scans of the underlying mapr-db table. These scans are very efficient in many ways (such as, aggregated, cached, and so on), but these are scans on disparate parts of the dataset. In addition, if the consumers are not caught up with the producers, these scans are against disk. This disk scanning slows the system down. With HDD, the seek costs might be quite substantial. With SSD, the impact is likely to be far less.
See the MapR user documentation for information about topic creation (Creating Topics).
Retrieving data ...