EC2 instance type selection: vCPu vs Memory

Hi there,

I’m trying to establish a cost vs performance set-up. My use case is 4Mbp/s video with 1 publisher and up to 50 subscribers. I recall reading a blog post where the suggestion was with OpenVidu an EC2 instance of type t3.large or t3.xlarge? T3.medium has half the memory but same vCPU count and is much more cost effective.

Is there a formula here I can use to try and figure out what capacity is needed.

So for example. When the video is transcoded for a single publisher it is only done once right? So in theory the question would be what is required to push up to 50 video streams to the subscribers. Is that memory and/or CPU and how much is needed? Does the fact that I’m running a front-end NodeJS application on the same server impact this at all and do I need to allocate more memory to handle the concurrent users on the NodeJS app. Or should I have a smaller instance and run this separately from the openvidu stack as openvidu takes care of pushing the video streams to the webrtc endpoint, i.e. users ?

Any suggestions here would be helpful.


Please see this post most of questions answers is there.
if there some doubt you have left soot it
Thanks
1 Like