Wowza Community

hardware requirement for 1000 live stream ?

Currently I am doing a research for building live video stream platform on AWS EC2 instances

and I need to know what is the best instances type for Wowza and how many instances do I need for 1000 live stream concurrently in HD 720p

We have an article on how to find the right server for your needs at the link below.

https://www.wowza.com/blog/4-tips-for-sizing-streaming-server-hardware

I have read the article but it is doesn’t say anything about concurrent inputs

and I found some benchmarks here

https://www.wowza.com/docs/wowza-transcoder-performance-benchmark

but there is something unclear about benchmark which is why if there is more than 10 inputs the output is limited to 10

Hi Abdalrahman, the benchmark you read about transcoding/transrating, if you dont need to transcode your 720p stream to the 480p 360p 240p 160p , then the benchmark is useless for you. If I’m not mistaken, In wowza the suggested live stream number is 100 at least 200 for per instance.

I can get 120 live stream in a single 32 gb ram 24core E5 2.x cpu with centos os.

You cant handle 1000 concurrent live stream input I guess, not only because of the cpu or gpu but also ethernet port limit & java limit . You need to distribute them.
if you need to transcode stream, I can give example from my ip cam system .
It use nvidia k2 grid gpu , run on centos os
input : 50 live 1080p stream and transrating to
480p,360p .

/dont mean to highjack the post, but i can’t send you a personal message on this forum.

Emre, can you tell me more about your transcoding setup. You are running 50 input streams 1080p @ ?? Mbps input and to how much streams can you transcode that?, also 50? Im curious about the nvidia cards transcoding.

What is your other hardware in that ipcam setup?

1080p at 4-6mbits ps, sony & axis ip camera. with 48 core cpu I cant transcode/transrate 50 concurrent ip camera. with nvidia k2 grid and 24 core cpu on centos. Only gpu transcoding cpu not used. I see the %70-80 gpu usage via nvidia-smi cli for both of card.