NDI failover speed

[rich99]

Hello,
We just started testing the NDI failover option when sending an NDI feed. If we crash the sender, the receiver will switch rapidly to the failover feed.
However, if we simply disconnect the ethernet cable, it takes about two minutes to switch.
Is there a way to make this more aggressive to have an almost instant switch or at least under 15 seconds?

Thanks

[livepad]

Hello,
We just started testing the NDI failover option when sending an NDI feed. If we crash the sender, the receiver will switch rapidly to the failover feed.
However, if we simply disconnect the ethernet cable, it takes about two minutes to switch.
Is there a way to make this more aggressive to have an almost instant switch or at least under 15 seconds?
Thanks

What you are seeing appears to be consistent with other testing. It seems likely that the low level TCP retry-before-fail mechanisms are responsible for keeping the socket alive whilst TCP tries its best to fix things before *finally* giving up after a number of seconds or minutes. It's only at that point that NDI is made aware that the socket connection has failed. Whereas when an NDI source *crashes*, the operating system actively closes the socket and the other end knows pretty much straight away

If you want to create a fault tolerant infrastructure which can failover quickly in the absence of NDI frames (regardless of the failure cause), then you may want to investigate the Sienna NDI Processing Engine. One of the 120+ modules is the Failover Module which can flip between primary and backup NDI Streams based purely on the arrival (or not) of NDI packets. The module also emits NDI Metadata which can drive other infrastructure turning things on and off, or reporting alarms etc.

http://sienna.tv/ndi/ndiprocessor.html

There is a short (internal training) video you can see here which shows it in action: https://drive.google.com/open?id=1wU...qb-c2I1OTv90F7

That first video shows the basic function of the failover module, but you can see a more sophisticated implementation here: https://drive.google.com/open?id=1u6...3UteLZlOpOb_FN

This second (internal training) video shows how the metadata emitted by the failover module can drive rich process automation - in this case controlling an NDI Router to create a salvo based response to the failover condition.