I’m wishing to use mediasoup to build the server side HA cluster, take video streaming for example, when real-time stream subscribers exceed the single SFU node’s throughput ability, a server-side tree-hierarchy topology needs to be built.
My question is, library mediasoup and its client is separated, the former has API for creating producer/consumer, but they are both passive, which means I have to build a lite client library to combined with the server mediasoup library, to build the streaming pipeline cluster solution?
Initially when i see the documentation, i thought the DirectTransport may be the answer, but possibly it’s not.
When saying “streaming pipeline”, i mean i can also put some AI-rendering or AR-effects to extend the simplest producer -> SFU -> consumer, not just for broadcasting HA. Or think about it as to put the gstreamer pipeline into the server side
The PipeTransport api has no clarifying test case, previously i’m only paying attention to DirectTransport with which to combine the streaming source and SFU node on the same host…
Both methods for scaling RTC producers/consumers across networks or CPUs. Examples aren’t overly given and if you need to understand how pipetorouter works, examine its code it’ll help you understand how pipetransports work.