Network UDP delay

[WatchDog]

WO version: 6.1.6

Production computer utilizing 2 NIC ports with separate IP range

1 Switch in network

Issue:

I have experienced cases where a display machine or group of display machines from a cluster react to a UDP command with a consistent latency. The latency can be anywhere from tenths of a second to 2 seconds. Jumping to a marker or scrubbing the timeline does not induce latency… only when running a play command from the production machine. That alone suggests it is something inherent with WO’s use of UDP commands. My theory is that there is timing information built into the UDP packet, and somehow either the timing information is incorrect within the packet… or on the Windows end, the UDP packet is being delivered late, and thus the timing information in the packet is offset to what the display thinks is the system time. Either way, it is a disastrous bug.

I have had it occur during heavy workload moments (transferring lots of data over the file share, making a high volume of changes to the timeline, etc…). On this last show, I didn’t remember to disable the file share port until I got into WalkIn. And the bug didn’t occur until we were about 80% into the show. But this is the first time I had the bug occur seemingly out of nowhere – as in I was not in a heavy production workload state when it occurred – just cueing through my show. Although I am not 100% convinced that it has anything to do with using a 2^nd nic, I still feel like the 2^nd nic being a root problem is a strong theory. I also have only encountered this when using virtual displays for video that then becomes re-entered onto the stage for distribution.

 

I am in the process of testing whether or not the active 2nd NIC port or use of virtual displays to map content/standard video playback to displays is responsible. In the mean time, has anyone else experienced this issue when using a control cue to execute an auxiliary timeline play command?

[Walter Soyka]

Yes, I had this issue on a show last year. Here's a thread about it:

http://forum.dataton.com/topic/2558-latency-on-running-cues/

 

The displays did have multiple NICs, and I did use virtual displays extensively.

 

w.

[WatchDog]

39 minutes ago, Walter Soyka said:

Yes, I had this issue on a show last year. Here's a thread about it:

http://forum.dataton.com/topic/2558-latency-on-running-cues/

 

The displays did have multiple NICs, and I did use virtual displays extensively.

 

w.

We have multiple NICs on displays as well, but the secondary ports are disabled on all display machines. Although not apples to apples, we both have 2 NIC ports on production computers with shows that contain VDs. Not to mention cueing aux's from the main timeline. Thanks for the info. Let me know if you find anything.

[Alex Ramos]

I'm using multiple NICs on production machine, and get lots of random network problems.

UDP is sent to both NICs TCP only to the one with the handshake ...

The only solution I found consistently working is to start Watchmaker only with the NIC for WO network ON, and after Online start the other NICs.

Even setting the NIC priority in windows doesn't work. 

But still sometimes I run into problems when updating the shows, as the handshake needs to be done again an Watchmaker not always grabs the correct NIC. 

We need an option in WO to lock to a selected NIC