Cluster Computing for Dedicated Server

JJRcop shared this feedback ago

Submitted

I posted this idea at feedback.keenswh.com earlier, and it didn't really take off when it comes to votes, but essentially this suggestion is listed there, here's what it looked like: 47d078f9eee8c4f5b875de49b105167c

Essentially this is a duplicate of the same suggestion, however after a comment thread on the feedback I posted before, I would like to clarify that this suggestion is to add this functionality to the dedicated server software, so that anyone can build their own server cluster, rather than just Keen hosted servers.

Since there was a multiplayer rewrite, I wonder if this is more feasible than before.

I like this feedback

Replies (3)

CheesyBread ●

Oh man oh man.

So I wrote up an idea on there that went into more technical detail on how this would work, but I cannot find it now and I forgot my login. In theory a system I designed (but only on paper, I have not written code for my own game using this idea) could allow for exactly what you are wanting, but I highly doubt Keen will ever do it.

Reply URL

Konstantin ●

And you can use a neural network to simulate physics.

Reply URL

VestedGamr ●

This would be awesome, but it's a MASSIVE project and I don't see Keen software ever doing this, especially seeing as though they're prioritizing DLC content over bugfixing, balancing or quality-of-life features. I'd suggest talking with some software engineers about making a 3rd party server program if you want to do something like this.

Reply URL

CheesyBread ●

That's a good idea actually. But sadly I do not know Keens code at all, and also their code isn't on Github like it used to be. Only their really old code exists while their new code is on a closed svn that only some select top modders have access to. Unless something new happened and I am wrong about that, I do not know how to obtain the source code.

Also I don't know their code base at all.

Lastly the architecture I designed that allows server clustering also requires the client to make and change connections frequently. In my architecture, the client has up to 16 connections (worst case scenario) that connect and disconnect frequently depending on what the player does. And in a more redundant version of my architecture, the client has up to 32 connections because each connection is given a backup connection to a secondary node which acts as a fallback in case one connection dies.

I do have an older iteration of my cluster concept that wouldn't require the client to be modified, but it wouldn't have as good of a latency as my best iteration of the cluster design.

URL