hard to configure the push speed when not knowing the current work load of consumers (some consumers might be busy with something, sneding too much messages to it may cause it to degrade in performance)
broker logic is more complex: where to send messages, how fast to send them, etc
RabbitMQ, ActiveMQ
Pull Type Message Queue
optimized for throughput: takes as many as necessary and waiting for consumers to digest
better suited for message replay
say if something goes wrong, and the consumer finds out 24 hours later, it can call the broker to replay all the messages in the last 24 hours
consumer controls the logic
polling in a tight loop
needs to constantly calling the broker to check for new messages, resulting in unnecesary resource waste
