[rabbitmq-discuss] clustering vs federation

[Michael Klishin]

On 18 Feb 2014, at 18:42, jayashree gn <jaishreegn at gmail.com> wrote:

> Was reading about how to use federation and was hoping to read or get more information about the advantages of using federation in production environment VS clustering.
> 
> Any pointers in this direction would be very useful in knowing pros and cos of our environment.  

Clustering is a form of HA: messages are duplicated between queue master and mirrors. Consistency
is favoured over availability (CP in CAP). See

http://www.rabbitmq.com/clustering.html
http://www.rabbitmq.com/partitions.html

Federated exchanges replicate messages that are published to them to another node (or cluster), asynchronously.
Availability is favoured over consistency: federated exchanges will reconnect and republish messages when downstream is not
available.

Federated queues distribute (but *not* replicate) messages between multiple nodes (or clusters).

So, they are not mutually exclusive and solve different problems.

MK

Software Engineer, Pivotal/RabbitMQ