Metrics to use to scale up servers


First of all, good job for all the work you guys provided to build this library.

I’m currently working on scaling servers to handle rooms with lots of consumer. After reading the documentation, and especially the Scalability page, pipeToRouter seems to be the way to go.

On this same page, you guys mention:

a mediasoup C++ subprocess can typically handle over ~500 consumers in total

Of course these 500 are related to the CPU and config of the host. But the question is : what is the limitation of the server ? In other words, what would we listen to in the server monitoring to scale up ? (Memory, CPU, files open, network ?)

Thanks !

There is no step by step guide. It depends on your machines and on your service. Whatever (Memory, CPU, network) exhausts first will be the indicator for scaling up.

You’ll have to monitor and see how the service works under load and what the load limit is for a machine to stop adding more tasks to it.

1 Like

I see, thanks @jmillan for the quick reply. So I need to do more testing in order to know when to trigger a pipeToRouter from the initial producers.