Downstream Servers


#1

@Frizzlefry ,

Can you or someone on networking please tell us more about this?

Scaleable to Crowds Support 100+ avatars together in one space, with broadcast options to reach 1000's with repeaters.

I have seen these options inside of the stack manager settings and I was asking about this a few years back. Seems as though this feature for multiple servers slid-in-under my radar. Others might want visibility on it also.

Please tells us about it.


#2

@AlphaVersionD

There is a doc here that explains how the “Broadcasting” feature works.

https://docs.highfidelity.com/create-and-explore/start-working-in-your-sandbox/broadcasting

I’m not 100% sure this is exactly the feature you are looking for - happy to take any follow up feedback/questions here.


#3

I was unaware of this document. :blush: however, I do have a few questions that will save me time. I have an assumption of what happens under this configuration, but I would be better served by asking:

Does a downstream audio-mixer SHARE the computational load of primary audio-mixer found on domain-server?
(stats result: servers = 7 and physical machines = 2)

-OR-

Does a downstream audio-mixer OWN the computational load of primary audio-mixer found on domain-server?

(stats result: servers = 6 but physical machines = 2)


#4

When you use the broadcasting feature you have separate domains where certain users are re-broadcasted from the upstream audio and/or avatar mixers to the downstream audio and/or avatar mixers.

So, an Interface client in each domain will see the same number of servers as a domain not using the broadcasting feature.

The downstream audio mixer will receive 0 or more audio streams for rebroadcasted nodes from the upstream audio mixer. Both the upstream and downstream audio mixer will perform the HRTFs for their listeners for the audio stream of the re-broadcasted node.

To the downstream audio and avatar mixers these rebroadcasted users basically appear as if they were directly connected clients. The computational complexity added by a re-broadcasted node is slightly less than for a normally connected node, because the re-brodcasted nodes are not sent any data from the downstream mixers.


#5

Being in 2 domains at once. Got it.

Thanks @b!


#6

I’ve tried this on many different machines. It isn’t working. I’ve attempted to split assignments to different blades, I originally tried doing link aggregation;
(for folks following along at home - http://www.pcgamer.com/how-to-set-up-nic-teaming-link-aggregation/)
but thinking that might be causing the problem, I got off my high-horse and tried just single pipes back and forth. Nada. I’m giving up on this until I can get some legit hand-holding because I’m convinced this isn’t working in the public beta release. Starting the domain-server from the cmd line is working (sort of) but no assignment client fires. Even using the -n7 on the specific domain-server box when done from the cmd line results in zero ac and therefore empty domain shown. Regardless if the assignment is locally executed or from a remote node. It just doesn’t matter. Here check it out:


For the sake of this experiment, please assume I have physical access to each machine, switch, and router participating in the test. Including 2 isolated stand-alone interface.exe clients if need be. I did read port teaming is not possible in Win10Enterprise, (but that doesn’t explain why doing single NIC failed to work… perhaps change the Link Agg Group on the smart-switch?) Learning as I go, and I know just enough to be dangerous. :blush:

Blades parked until you can help me sort it all out…

Thanks @b. Happy Holidays.


#7

Let me know if you believe I’ve done something adverse to your recommendations. I’ve tried following the instructions as closely as possible. No dice.

cc: @Frizzlefry