[rabbitmq-discuss] 2 nodes, 4k queues

[Matthew Sackman]

On Fri, Nov 20, 2009 at 03:08:23PM +0000, Gojko Adzic wrote:
> > inappropriate messages in the consumer. I know this isn't ideal, but it
> > might solve the problems you're seeing.
> 
> direct or fanout is not an option as we'd burn the bandwidth. is header 
> matching more efficient(we can do the same with header matching)? we 
> have ~4 bindings per queue.

Header matching is at least as bad, if not worse. We're only talking 16k
bindings though, I'm wondering what is the bottleneck. Are these
messages persistent and being written to disk at all? Do the queues get
particularly long or stay pretty much empty?

> > this case. What is your topology on the publishing side?
> 10-20 publishers running in parallel.

Hmm. I'll do some testing, but I'm wondering whether something else is
going on here...

Matthew