The optimum number of MQ servers in any scenario depends on the expected number of clients that will be connected at any point of time and distribution of producers and consumers depends on the JMS connections and destinations.
As the number of producers/consumers per connection increases, there will be different throughput observations. For example, having 1 publisher and 1 subscriber per connection in an n:n scenario will give better throughput rather than having more than 1 publisher or subscriber per connection.
Similarly there is a limit to the best throughput that can be extracted from the server by increasing the producers and consumers on a destination.
For example, the throughout of the FioranoMQ Server will be higher for cases where there are 10 producers and 10 consumers on a topic but the same will drop if this figure is increased to say 100 because now 100 clients will access the same destination in-memory buffer leading to slow processing of each request.
The in-memory buffer sizes explained in the section before this also govern the above mentioned performance changes.
From the above explanation it becomes clear that if the deployment scenario of FioranoMQ involves a large number of client applications (> 500) sending and receiving messages at high rates on a large number of destinations the throughout results can be increased by distributing the load of both the client connections and the data amongst different instances of the FioranoMQ Server.
The number of server's instances to be used depends solely on the number of JMS connections created to the server and the number of destinations used on the server.
In a sample scenario where there are about 100 producers and consumers each on about 500 destinations the most optimum throughput would be achieved by sharing the load across 2 FioranoMQ Server instances though one FioranoMQ Server would also be able to take the load however, might not give the desired throughput due to memory and CPU limitations. The administrators can divide 250 destinations on each server for better throughput.

Adaptavist ThemeBuilder EngineAtlassian Confluence