Ridiculous amount of bandwidth being used by clients since update.

Hi all,

I am seeing a huge spike in upload bandwidth since the most recent update was applied. The source media has not changed but it seems like a single streaming client can peg my 15 Mb/s upload easily. The media being streamed says it’s being transcoded.

Previously I was able to stream to 5+ clients simultaneously while keeping bandwidth usage under 12Mb/s, so something has changed with PMS.

Is there a way to limit the upload?

The reason your upload is getting pegged is because the clients are calling for more. This stems directly from their Plex/Web Player settings.

Since there is not yet any way to limit upload bandwidth, the only thing which can be done is to do the math yourself and have them make the adjustments. to their playback settings.

Right, but the clients are not Plex/Web clients. The clients are primarily AppleTV/LG WebOS. I have confirmed the streaming settings on the devices is set to 4Mbps, but am still seeing the bandwidth peak.

Is there a way you can take one of those clients and capture a playback/transcode session with logging set to Debug (not Verbose) or do you have such a log ?

I would like to see which profile(s) are being used. Also, which version of PMS are you running and on what host platform?

I’ll see if I can capture one of the logs. It may have to wait until tomorrow.

I am running PMS v1.0.1.2396 on Ubuntu Server 14.04.1.

See the PlexMediaServer.log attached. Client started stream, confirmed setting to 4Mbit 720p and I observed about 10Mb/s upload speed. The specific client tested for this log was “Plex for LG (webOS 3.10.19-32.afro.5)”

Thanks for that log. It does show (search along with me to follow):

  1. The source is about 12 Mbps.
  2. The transcoder is invoked with a 4000k limit (MDE result of client negotiation)
  3. The average upload is about 4.03 Mbps, at times dipping to 3.97

I suspect you’re seeing what we all see; a flurry of activity while the buffer fills then nothing. Playback streams along from the buffer until it detects it needs to be filled again. At that time, you get a full-speed upload to fill the buffer. The cycle repeats until playback halts.

This is the type of QoS flow control needed. It is a long-standing request and not something the devs have had time yet to implement (prioritization of resources / people).

I can only suggest you seek it out in the Feature Requests forum and vote up the need for it. The most wanted features there get the best shot at making it into the ‘do now’ list.

Thanks for having another look at the logs. I don’t feel like I am going so crazy now.