On Fri, Oct 9, 2015 at 4:03 AM, Benjamin PETIT <
petit.b...@gmail.com> wrote:
> Thank you all for your replies.
>
> So if I recompile OLA I can get more universes in SACN.
>
> I will be doing some testing publishing to 128 universes even if I just have
> one node on the other side.
Let us know what you find out. The last performance numbers we have
are from many years ago.
There is a change we can make to the RPC protocol which would reduce
the number of memory copies.
>
> I understand that if I want to optimize latency and framerate I should get a
> IGMP snooping switch to get the most out of multicast.
> - would you recommend a specific switch (16-24 ports) that work well (not
> too expensive)?
> - what kind of configuration do you need to do? I am not a network expert.
I'll leave others to comment on that.
> How does it work?
> - For artnet if I understand well the different nodes publish the universes
> they specifically handle, put this is limited to 4 nodes/universes?
Correct. A node with 1 IP address can handle up to 4 universes. There
is a trick to support more than 4 but that requires multiple IPs per
node.
The OLA artnet plugin can send up to 16 universes of ArtNet. The
limitation is there because of how we map OLA universes to ArtNet
universes.
> - For SACN I don't understand how the packet redirection works on the
> switch. Is it part of the redirection?
sACN uses multicast:
https://en.wikipedia.org/wiki/IP_multicast
>
> Is there a way of doing true unicast in Artnet or Sacn without having
> multiple network cards? Imagine I have 12 nodes, each of them handling 16
> universes without overlap between nodes.
> If I know the pairs Node IP / Universes handled, could I just send to each
> IP the DMX data concerning the correct universes?
Yes. You don't need multiple NICs for that.
Simon
> Thank you again all for your help. I worked quite a lot on small artnet
> projects (up to 4 universes) and I thought it would be easier to switch to
> bigger project.
>
> Benjamin
>