Choosing TURN server (like Coturn) vs Building Custom Application for WebRTC

[Mike Shackleford]

Hello....In the past, I've developed simple Red5 applications for RTMP, extending the MultiThreadedApplicationAdapter and registering PublishInterceptors and PlaybackInterceptors to integrate into the rest of my software stack. Worked pretty good! But I'm definitely not an expert and I'm getting hung up on one thing. 

I have a new project that requires an attempt at WebRTC peer-to-peer using STUN, then falling back to some sort of relay if that doesn't work. Ant Media seems to have some nice features I'd like to explore, but where I'm caught up is when to use an Ant Media application for such a relay, or when to install a TURN server (like coturn) and use that instead as some sort of relay. Is one more efficient than the other? Is one (perhaps application development) just much more feature rich? Or have I completely missed some fundamental concept (known to happen from time to time :-)). 

My gut feeling is developing an application would be more flexible, I'll need to do some restreaming, some Adaptive bitrate stuff for sure, maybe cloud storage, etc. Is anyone willing to share any general thoughts in this area? Anything is helpful to me. 

P.S. : I'm lucky in a way as I have full control of the whole system (hardware, mobile, cloud, firmware) so I have a lot of flexibility, but I sure don't want to pick the wrong thing. 

[Connessione]

@Mike Shackleford the ecosystem fir AMS is somewhat different from what you described having had with other products. Here things are pretty much crafted to specific solutions. There is isnt much to do on server side because mostly everything is done either though rest apis or client side.

for example the security is implemented through jwt so that a lot of people who might not be developers or java friendly can get it working as a standard. A server application would be more flexible but the demand to go that fine grained is much lower than just getting security working via something like JWT.

That said it is definitely possible to build a server app in java as well but you will need lotS and lots of hand holding from AMS team as there is documentation on the java apis.. It is just better to changing your thinking from server oriented to client oriented in this case.

 

going over your other post I understand you needed serverside development options for using AMS as signalling server purely in some cases. So in this case I will suggest exploring nodejs or something as a efficient option. When you need be to many or many to many or even one to one with recording...then use AMS.

[Mike Shackleford]

Thank you so much for taking the time to provide this insight. My application is going to require deep, broad integration with a much larger cloud stack that has its own authentication mechanisms, notification systems, etc. Our streaming section of our platform, although significant, is just one small part of a far bigger existing platform, so custom integration is unavoidable in my case. 

It's good to know that such integration is possible, even if it is difficult and not the standard way to do things in the AMS world. I'm learning! Thanks again. 

[Connessione]

Just repeating my sentence here which got mistyped earlier:

That said it is definitely possible to build a server app in java as well but you will need lotS and lots of hand holding from AMS team as there is not much documentation on the java apis.. It is just better to changing your thinking from server oriented to client oriented in this case.