[rabbitmq-discuss] Is the C client a show-stopper for rabbit?

[Alexis Richardson]

Jiri,

On Tue, Feb 15, 2011 at 1:52 PM, Jiri Krutil <jiri at krutil.com> wrote:
> Alexis
>
>> We are certainly willing to improve the C client and it is something
>> that is on the "todo" list.
>
> That would be great.

Indeed.  See David's comments.


>> Yes there is a C++ wrapper which is here:
>> https://github.com/akalend/amqpcpp
>
> Will look at it. Did you consider referencing it from
> www.rabbitmq.com/how.html#clients ?

Good point.


>> In relation to performance, as always it comes down to what your use
>> cases are.  The Rabbit implementation is quite efficient, in most
>> cases but not all.  What kind of scenarios did you have in mind?
>
> Our use case might look like this:
> - mostly pub/sub messaging with topic exchanges and word-based binding key
> matching
> - few producers and hundreds or thousands of consumers using the same
> exchange
> - no need for persistency or delivery guarantees, no transactions
> - mostly auto-acking messages
> - message payload usually less than 2 Kb
> - relying on order of message delivery to a large extent (but never reject
> or not ack messages)
> - we need low latency

Understood.

I have three questions:

How low latency?
What ingress rate?
Would pubsub be broadcast (direct/topic exchange) or anycast (shared
queue)?  I assume the former...

alexis