[zeromq-dev] architecture

Pieter Hintjens ph at imatix.com
Tue Jan 25 00:10:52 CET 2011

On Mon, Jan 24, 2011 at 11:54 PM, Benoit Chesneau <bchesneau at gmail.com> wrote:

> Yes, but I may have missed something. So far I was under the
> impression I needed a broker Where I 'm trying to avoid that. But I
> may be wrong, back to the doc then, thanks.

Brokers can simplify architectures with many peers.  They solve some
problems such as fair distribution from N clients to N workers. But
you can also wrap up broker-like functionality and embed it in each
client, if your architecture allows it.  ZeroMQ does this nicely,
given it's interthread communications.

I'd suggest that you do put brokers into the architecture where it
makes sense at first, and you can then later optimize them out where
they become bottlenecks.  Avoiding brokers just for the sake of
avoiding them is generally counter-productive.

Pieter Hintjens

More information about the zeromq-dev mailing list