ejabberd - max number of participants in MUCs vs. PubSub

Question

I am building a mobile chat platform with support for both private and public rooms. For private rooms, I have already successfully implemented the MUC functionality using ejabberd on the server and XMPPFramework for my iOS client. However, for public rooms, I would like to evaluate whether PubSub makes more sense or not?

To define the use case in detail, all public room members should be able to post / send messages, as well as have offline messages delivered to them when they login. The rooms should be persistent as well. Also, I would expect the max number of participants in a public room to be many thousands (eventual state if I were to plan for high scalability right now). This would differ from the private rooms where I intend to set the max number of participants' limit to be 200 or so. For the private rooms, I am exchanging presence, but for the public ones, presence exchange is not mandatory if that overloads the server.

Can folks advise please what should be the approach recommended here? Also, are there any pointers as to how I can go about capacity-testing these approaches...

Many thanks!

Answer 1

For benchmarking ejabberd and comparing performance, you can use Tsung: https://github.com/processone/tsung

It should help measuring yourself the difference of behaviour for your user case.