Earlier, during the explanation of our system architecture, I briefly discussed the possibility of fanning out messages from the stream listener to multiple queues. A design such as this would be useful when there are different types of workload to be performed from a single data producer. The following example architecture shows a system made up of an individual Twitter stream data producer that fans out messages to multiple queues based on the payload:
For this example, assume we're interested in processing a more extensive range of tweets. This system will still classify dog and cat images ...
