[rabbitmq-discuss] RabbitMQ with WAN latencies
alvaro.saurin at gmail.com
Tue May 17 14:59:54 BST 2011
On 17/05/2011, at 15:28, Simon MacMullen wrote:
> On 17/05/2011 12:39PM, Alvaro Saurin wrote:
>> I was wondering if you could help me determining if RabbitMQ is
>> appropriate for some WAN scenario...
>> I have to send control messages to 50 or 60 machines deployed around
>> the world, so they can subscribe to some message types and then
>> perform some actions. Maybe some machines will be grouped in small
>> clusters of 4-5 machines, and some tasks can be run by any machine in
>> that group, but there will also be some tasks that will be targeted
>> to some specific machines. There should be a very low traffic, with
>> maybe just a couple of messages per second, so I was planning to use
>> a simple architecture, with one central broker.
>> I want to keep things simple, but I'm a bit concerned about the
>> effects latency could have, and I have also read that connectivity
>> losses could have some negative impact. My question is, could I keep
>> this simple architecture or will I need something a bit more
>> complicated like rabbitmq with xmpp or shovel? I have not found many
>> docs about using RabbitMQ on WANs, so maybe nobody is crazy enough
>> for trying this... maybe I should not use RabbitMQ at all? :)
> If you are going for a central broker model you should be fine. The discussions that we've had about network partitioning and use of the shovel to mitigate it have all been about clustering over a WAN, which is not recommended. But AMQP over WAN is fine.
> Obviously synchronous methods will be slow when latency is slow (although no worse than e.g. HTTP would be), but with low traffic like that you probably don't even care.
> Cheers, Simon
> rabbitmq-discuss mailing list
> rabbitmq-discuss at lists.rabbitmq.com
Great! That´s what I wanted to hear! I´ll try this setup and see what happens...
More information about the rabbitmq-discuss