Configuration WCS livestream up to 300 streams ON AWS EC2

[Gain]

Hello.

I use WCS on AWS service, And I have one question to ask.
How can I config WCS on AWS service to livestream up to 300 streams in one time ?

Also, I want to know recommend package for up to 200 livestreams that I should use for WCS on AWS service.

Please advice.

Gain

 

[Max]

Good day.
Seems m5.2xlarge should be enough. Please also make a tuning steps described in this article, chapter Preparing for testing, points 1-5

 

[Gain]

Good day.
Seems m5.2xlarge should be enough. Please also make a tuning steps described in this article, chapter Preparing for testing, points 1-5

Thank you for advice.
I have one more question to ask.
I use WCS on AWS service, do I have to buy license too or it already include with it ?

 

[Max]

I use WCS on AWS service, do I have to buy license too or it already include with it ?

If you're using AWS Marketplace AMI, you have not pay additional charges, just hourly payment for instance.
If you're deploy Amazon Linux for example and install WCS manually, you should buy a license.

 

[Pin]

Hi Max,

I and Gain had used flashphoner on produnction. I attached cpu usage from AWS on follwing.

Here is our enivornment and use cases.
1. We have up to 600 web cam steaming to 14 instances and devided 40-50 each instance.
2. We have maximun 195 view sessions to view 600 steaming each time (both audio and video).
3. We use m5a.2xlarge on AWS
4. We have done setting up heap size configuration.

Here is our question and requests.
1. I think it seem too high cpu usage (70%) is that normal for 40-70 steamer? and not match with your answer that m5.2xlarge should be enought for 200 steamer.
2. Why? Sometime, the steaming is lost and user have to establish new streaming again. For 2-3 hours steaming.
3. By early of next month, We have to handle 15,000 steaming connection. and 150,000 stramer by end of next month.
4. If I use above enviroment for 150,000 steamer, it will be amount 3,000 of m5a.2xlarge. I think it seem impossible to do it on AWS and it will be a huge cost.
5. Can you providing a kind of service to us for config the instance and load test together with our team? Please.

 

[Max]

1. I think it seem too high cpu usage (70%) is that normal for 40-70 steamer? and not match with your answer that m5.2xlarge should be enought for 200 steamer.

Please check if stream transcoding is performed on the server. If the following parameters on the statistis page http://server:8081/?action=stat are not zero, then some streams are transcoded:

native_resources.video_decoders=0
native_resources.video_encoders=0

For transcoding, 1 CPU core is required for 2 720p streams encoding, or for 3 480p streams encoding. So check what resolution/codec do you publish, and what resolution/codec viewers using to play. You can check stream metrics using REST API or statistics page http://server:8081/?action=stat&format=prometheus&groups=publish_streams.
The streams should be played using the same codec and the same resolution to prevent transcoding.

2. Why? Sometime, the steaming is lost and user have to establish new streaming again. For 2-3 hours steaming.

Please note that server should be specially configured to handle a big subscribers amount. Read this article What kind of server do you need to run a thousand WebRTC streams? for details, section Preparing for testing.

3. By early of next month, We have to handle 15,000 steaming connection. and 150,000 stramer by end of next month.
4. If I use above enviroment for 150,000 steamer, it will be amount 3,000 of m5a.2xlarge. I think it seem impossible to do it on AWS and it will be a huge cost.

If you need more than 2000 subscribers per server, you have to deploy a CDN. Please read this article for step-by-step deploying example, and this doc for technical details. And, for 2000 subscribers, it should be more powerful config than m5a.2xlarge (m5a.16xlarge and higher for example).

5. Can you providing a kind of service to us for config the instance and load test together with our team? Please.

You can request a presale. In this case, to make a load test, you should provide at least two servers: one to test and another as testing server. Please note that testing server must be more powerful than server to test, at least twice.