It depends on the kind of bottleneck. Is it the network? Is it the cpu?
memory?
One could simply keep both producers and consumers together and just have
many producers/consumers.


2013/5/8 Pieter Hintjens <[email protected]>

> You're asking two questions, one about what pattern to use (push-pull
> is fine), and two about performance. Cost is always relative to
> consumption. So your serialization may be fine, or may be a bottleneck
> (it's not a 0MQ issue), and you can always make it faster if you need
> to. You cannot avoid serialization however unless your two tasks are
> in the same process, and you pass pointers around (over inproc).
>
> -Pieter
>
> On Tue, May 7, 2013 at 11:17 PM, Stathis Gkotsis
> <[email protected]> wrote:
> > Hi all,
> >
> > I have two processes: one producer and one consumer.
> > The producer does some processing and builds objects, while the consumer
> should take these objects and perform further processing.
> > The idea is that these two processing tasks work in a pipeline, since
> they are independent.
> >
> > I think that ZeroMQ could be a good fit, I could, for example, use a
> PUSH-PULL socket for the two processes to communicate.
> > The problem is that, it is not only messages (text) that the two
> processes need to communicate with. It is big objects that need to be sent
> and I would like to avoid any costly serialization.
> >
> > Any ideas?
> >
> > Thank you,
> >
> > Stathis
> > _______________________________________________
> > zeromq-dev mailing list
> > [email protected]
> > http://lists.zeromq.org/mailman/listinfo/zeromq-dev
> _______________________________________________
> zeromq-dev mailing list
> [email protected]
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>



-- 


Sincerely yours,

     Apostolis Xekoukoulotakis
_______________________________________________
zeromq-dev mailing list
[email protected]
http://lists.zeromq.org/mailman/listinfo/zeromq-dev

Reply via email to