[rabbitmq-discuss] Architectural Choices

[Michael Delaney]

I've got a web app on ec2 and I'm introducing async messaging with rabbit. 

My goal is more towards *reliability* than for maximum performance. 
Having said that, what are the implications of the following strategies:

Given: *3 Clustered Web Servers, 2 Clustered App Servers*

The scenario is that I'm publishing events from the web servers, which will 
are subscribed to by services running on the app servers. 

*Option 1*: Run a broker on each in a HA Cluster, sharing the queues 
between web & app servers. 
*Option 2*: Run a dedicated messaging server which all the web & app 
servers connect to 
*Option 3*: Run a broker on either all the web servers or all the app 
servers

At this stage I'm going with Option 1, which I feel is more difficult from 
an operational sense, i.e. managing configuration etc, but should give 
me the most reliable setup. 

Option 2 would probably be easier to get up and running, but I'd need an 
extra server or 2, and if the server goes down (it is the cloud after all)
i'll end up with service down time. 

Option 3 is basically somewhere in between. 

Is there anything else I need to consider?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/attachments/20111216/5eba4926/attachment.htm>