Amazon g2 instances for encoding

Community Forums Backend Server Support
1

Hi

I want to use Amazon EC2 G2 instace for kaltura encoding.

I am wondering if it will be enough to compile ffmpeg.
I think there also should be some configuration in Kaltura to use the GPU for encoding.

anyone have an idea?

2

Hi @benny_1,

If I understand correctly, you want to compile your own version of ffmpeg to utilise GPU acceleration? as described here:
https://trac.ffmpeg.org/wiki/HWAccelIntro
?

If so, note that I cannot assure you this will work correctly as we only officially support the FFMPEG RPM/deb shipped with CE. However, you are of course more than welcome to try.

It is important however that you use the RPM/deb spec files as a base, you can see them here:

It will also help you understand which dev packages you need in order to do so.

As far as compiling it on a G2 instance, what exactly is the HW spec for this Xen image? I think any machine can do it, it will just take a lot longer on weaker machines.

There is also this guide:
https://trac.ffmpeg.org/wiki/CompilationGuide

Hope this helps,

3

The problem is not the compilation, but how to make kaltura use it the right way.

The GPU encoding is being done if you use the nvenc encoder and not libx264

where can I change the encoding parameters passed to ffmpeg?

4

Hi,

With emphasis on the fact we will not be able to provide official support should this malfunction, and also that you’ll need to reapply whatever patches you make at each upgrade, the relevant code is here:
/opt/kaltura/app/infra/cdl/kdl/

For example: /opt/kaltura/app/infra/cdl/kdl/KDLOperatorFfmpeg1_1_1.php

5

thanks

which one of these files is actually being used for ffmpeg encoding?

6

Hi Benny,

Do you have best settings for convert use GPU with Cuda?

7

Sorry,
Could not find any configuration that works yet.