Hello, Wowza Wizards. Here’s a neat one:
ARCHITECTURE:
We are publishing live streams (coming from both a HaiVision Barracuda and a Makito encoders) over to a Wowza for EC2 Origin/Edge config, from very different geo locations, and with all sorts of possible combinations of resolutions, bitrates, framerates, GOP and what not. We then play back the stream using JW Player and the included test/demo player.
The source we are using is interlaced video coming into the encoders.
We are streaming from the encoders using TS over RTP, and Wowza is properly configured as it always has been (it’s been working fine until recently, unless we somehow missed those annoying pauses).
ISSUE:
In every case, we are noticing a small “break” / frame skip / pause in playback at a very precise interval - let’s call this interval “n”. We have determined pretty certainly that n = GOP / framerate, and we can reproduce the behavior pretty simply, and predict how often pauses come based on this value. So, let’s say we have a stream running at 30 fps and a GOP of 90, we’ll have one of those breaks / pauses every three (3) seconds.
Curiously yet: by accident, we discovered that if we de-interlace the video (using the encoder’s built-in de-interlacer) ONCE the player is connected (that is, while we are viewing the video) and the switch back to interlaced video then WHILE WE ARE STILL CONNECTED to the stream, the playback loses the annoying pauses; yet, if we re-connect, refresh or pause the stream, we go back to the previous state - that is, the annoying pause.
We are pretty confident this is not a player issue, since it happens with ANY player we tested, and we really doubt the issue being the encoders, since until recently they worked fine (although we also opened a support ticket with HaiVision to see if they can help). Also, the network shouldn’t be the problem here, since the behavior occurs over very different network scenarios, regardless.
We’ve been scrounging the forums trying to find some adjustment/solution/workaround, to no avail. We have implemented every possible adjustment suggested, but we haven’t been able to “cure” this.
Any suggestions? Do you need additional information to help us diagnose this situation? We can provide you with a link (although not a publicly displayed one right now) where you can observe the behavior if you deem it necessary.
Thank you very much in advance.