Performance on QNAP : Docker VS qpkg (Transcoding ok)

[GTsitouridis 2]

I have an issue with docker on QNAP in terms of performance.

While I have setup transcoding and it is working just fine, I have a reduced bitrate and as a result bad image.

The source is the same (TVHeadEnd on Pi Zero).  When using qpkg I get a bitrate of 20Mbps+ while when in docker I get only 10Mbps. At least this is the only thing I can see as a difference. I am also previewing the same channel while switching from docker to qpkg.

I did not know if it had to do with networking internal of QNAP, so I tried all possible docker connection methods, all of which yielded the same results.

 

Any ideas?

[Luke 37025]

Hi there, can we please look at example sessions from each so that we can compare?

Thanks !

[GTsitouridis 2]

Would you like me to sent over some screenshots or sort videos?

[Luke 37025]

Can you please provide the logging information requested in my link above? Thanks !

[GTsitouridis 2]

4 hours ago, Luke said:

Can you please provide the logging information requested in my link above? Thanks !

Hallo,

So I repeated a test and got logs, image sample and dashboard transcoding info. 
qpkg test from 10:02 to 10:05 local time

docket test from 10:10 to 10:13 local time

Logs_docker_10-10_10-13.zip Logs_qpkg_10-02_10-05.zip

[Luke 37025]

You're getting basically the same transcoding performance from both. I wonder if it's due to the disk or network performance of the docker container.

[GTsitouridis 2]

1 hour ago, Luke said:

You're getting basically the same transcoding performance from both. I wonder if it's due to the disk or network performance of the docker container.

Well.... both are running on the same physical system and on the same mirrored disks.

[Luke 37025]

Right but Docker adds a virtualization layer in between. That doesn't necessarily mean it will make a substantial difference, but there is a cost to that.

[GTsitouridis 2]

2 minutes ago, Luke said:

Right but Docker adds a virtualization layer in between. That doesn't necessarily mean it will make a substantial difference, but there is a cost to that.

This is more than true. I am just reminding that we are talking about the same physical system and the same array/volume.

One other thing, is that I have allocated all available resources to the container, which is barely using them.

Edited January 28, 2021 by GTsitouridis

[GTsitouridis 2]

On 1/28/2021 at 9:16 AM, Luke said:

Right but Docker adds a virtualization layer in between. That doesn't necessarily mean it will make a substantial difference, but there is a cost to that.

Hallo,

Eventually I think I figure out as to why it is happening. Comparing HW detection logs I found that native emby package on QNAP uses i965 driver, while in docker (running on QNAP) it is using iHD driver. I think this is the case with the following thread:

Hardware Encoding Issues (j4125) - Asustor - Emby Community

I simply do not know how to go about and correct it.

[Luke 37025]

OK that's interesting, thanks for the info.

[GTsitouridis 2]

Just now, Luke said:

OK that's interesting, thanks for the info.

In some thread you have proposed the use of the beta version, which worked ok for me too, thought the image was a bit soft, but reported as HW transcoding.

For a week or so, I have been using ghcr.io/linuxserver/emby and that is working perfectly!

For this container i965 driver is used.

If you wish, I can provide logs.