On 07/02/12 22:20, methodin wrote:
So when I was originally mapping out how to design an implementation
of RabbitMQ I assumed that an exchange could route messages to queues
and workers could consume solely off one of the queues. When I added
two queues in I found out that the message will get sent to both
queues and those both workers.
So I implemented a topic exchange and route to queues based on routing
keys. I am now, however, at a loss in figuring out the point of the
queues in this system as I could just as easily map 4 different keys
to the same queue vs. routing each one to their own queue. Is there a
large difference here? Should I be aiming for one approach vs the
other?
So we're comparing one topic exchange with multiple routing keys all
feeding into one queue, versus one topic exchange with multiple routing
keys each feeding into a separate queue?
There are a couple of differences which may or may not matter to you:
1) If more than one of your routing keys match a given message, that
message would get duplicated to each queue in the multiple queue case.
2) By pushing all the messages into one queue you impose a total
ordering on your messages, which you may want. But by splitting out the
messages into multiple queues you may get better performance (if the
performance of the queue is a limiting factor...)
Cheers, Simon
--
Simon MacMullen
RabbitMQ, VMware
