My setup is very simple, producer, consumer and activemq are all running on the same machine. If messages are relatively small (<100K), I don't see any noticeable delay. However, as message size increases, the delay between enqueuing a message and delivering it to the consumer grows exponentially. My average message size is about 1M and I see ~ 2 minutes delay between enqueuing and delivering to the consumer. There are no exceptions or warnings in the log, no CPU spikes, no GC cycles, nothing like that, everything looks normal.
If I change protocol from nio+ssl to any other (plain tcp, plain ssl or plain nio), messages are delivered instantly.
This behavior is reproducible with ActiveMQ 5.9.0 and 5.9.1 on OS X and Red Hat, JDK 1.7.0_45 64-bit.