I need to set up a new Wowza instance on EC2, which will need to take 50 incoming live RTMP webcam streams simultaneously (only 400kbps each), with client (viewer) connections likely to only be up to 10 max per stream at any one time (more often than not it’ll be less or 0). Live streams will be viewable on a website, but also need to work over HLS for iUsers. Not a particularly demanding scenario, but I want to get it right straight up. Can anyone recommend which AMIs would be fine for this? Would m3.medium be up to the task, or do I need to go bigger?
Any advice on this would be much appreciated.
Hi,
There’s some useful data in this thread.
The info there is about two years old, some of the posters there have done their own tests, but still your mileage may vary, and with 50 live source streams you really should start towards the upper end of those AMIs, as incoming streams do use up considerably more resources in comparison to outgoing connections.
Daren
Hello,
Both CPU are important. However, the maximal recommended value for the Wowza Java Heap Size is 10GB. So, if you are planning to run a single Wowza instance on your machine, you won’t need more than 16GB of RAM memory.
However, if you are planning to transcode the incoming streams, you should consider using an instance type with a lot of CPU resources. To get an idea on how much CPU resources are being used when transcoding different number of incoming streams, you should take a look at the “Wowza Transcoder performance benchmark” forum article.
Zoran
Hello,
Correct, HLS would be considered different delivery/streaming method. Not a transcoding process.
Best regards,
Andrew
That’s handy, thanks Darren, I’m currently looking at the large and xlarge c3 and m3 options. Can you offer advice on whether CPU or RAM is more important for handling incoming RTMP streams?
Thanks Zoran, there won’t be any transcoding going on, the input streams will be delivered on the web exactly as they arrive (H.264 640x480 25fps 400kbps). HLS will also be used for iUsers, but that’s not a form transcoding I believe? Just a different delivery method?