[zeromq-dev] Is there already native "high level" connection pool?

[Michel Pelletier]

On Mon, Nov 26, 2012 at 1:05 PM, Stefan de Konink <stefan at konink.de> wrote:

> On 11/26/12 20:19, Michel Pelletier wrote:
> > Can you explain a little more why you want to do this?  Why not just
> > have one socket per thread?
>
> Concept is:
>
> webserver           pubsub         client
> HTTP Post => PUSH   PULL => PUB    SUB
>
> HTTP Posts come in in huge chunks and get processed by about 1k for each
> iteration.


1k of what?  Iteration of what?  There's not really enough detail to
understand your problem.


> Multiple request at the same time would block eachother
> within the same thread. This is not solved by allocating a socket per
> thread, because that cannot guarantee that the incoming post is
> sequentially written on the same socket.


How would your pool proposal provide the same guarantee?  It sounds like
you should just PUSH your posts on down the pipeline as soon as they
arrive, and let a downstream consumer deal with accumulating them into a
single multi-part message.


> Data from a different source
> might be available prior to the previous request being finished.
>
>
> Stefan
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20121126/87c2f83d/attachment.htm>