Latency compared to Yggdrasil?

[lonnietc]

Hello,

I have been investigating Yggdrasil for possible use in a project and noticed that it does not self-organize or reoptimize to try and keep the network in optimal condition with near neighbor nodes.

When searching more, I came across your RiV-mesh which it seems to do these node re-optimizations and is also a fork of some previous version of Yggdrasil as well.

Can you please tell me more about how the latency changes relative to Yggdrasil in your RiV-mesh version since I need the highest speeds possible for a project that I am doing and wanted to try and get a bit of information from you on this if you had any.

Thanks and have a great day

[vikulin]

@lonnietc could you please elaborate a bit more what parameters are you are interested in? Latency meaning is pretty vary across different conditions.

[lonnietc]

Hello,

Thanks for moving this over to a discussion.

For a very long time now, I was thinking about forking the latest Yggdrasil and implementing a self-adjusting self-optimizing approach that would minimally optimize against neighbor latency and jitter ( plus some other network metrics).

The idea here is that a new node joins the network from any public IP and is then connected to a Ygg peer that may/may not have a public IP but does have a Ygg IPv6. Since nodes can connect and ping to any other node in the network once connected, this offers a great opportunity to use something like "yggdrasilctl getpeerremote IPv6=..." that will return the peers of remote node along with the RTT from the remote node to its peers.

For example if the local node hss 4 peers and each of those peer nodes have 4 peers then we would be able to get a list of 16 near-neighbor peers. The local node can then ping for RTT/Jitter/ etc. against the remote peers and thus have a list of 16 peers sorted by lowest metrics. The local node can then take the top 4 (i.e. lowest RTT) and use "yggdrasil addpeer IP=..." and "yggdrasilctl removepeer IP=..." to update its current peer list. Each node in the mesh network can do this independently as needed and we would eventually arrive at an optimized RTT mesh network that would dynamically adjust and update at the peer level as needed.

Of course, I would have to develop and embed a couple of functions within the daemon/service server so that it optimizes over some time period, but the effect should be to have a highly optimized network.

The rationale is that I want to have a high-performance P2P network that will be used for a number of P2P applications as well as graphic systems/servers and need the best performance that can be achieved for what I am calling the "Nexus Universe" (although that name could change)

I would be interested in knowing what RiV Mesh uses as the self-adjusting approach for your nodes and also perhaps we might be able to discuss how we might be able to collaborate on seemingly similar efforts.

Also, would it be correct to assume that you took a fork of Yggdrasil v0.4 as well given that your mesh is v0.4 just as a guess? If so, then are you planning to push your changes against the latest v0.5.12 release?

Best Regards and have a great day