The Word of Lordhood

lordhood117 wrote:

It has recently come to my attention that
xKillerbees, owner of the server Industrial Rage (the #1
IndustrialCraft/Buildcraft/Redpower server), has been trying to fix the
large amounts of lag on the server. His idea on how to do it sounds a
bit like what this mod is trying to accomplish. Since the IR server has
multiple cores, the plan is to run several servers together, each
linked via teleporters with inventory persistence and the ability to
chat between each world. Personally, I see an opportunity to work with
Killerbees to build a fully-working multiple world "super-server," which
I believe would be a major milestone in the development of this mod.
If it works, that could mean quite a bit of publicity for Futurecraft.
Please consider my idea here and let me know if you're interested.


Sincerely,

lordhood117

What do you guys think?

Publicity? Hell yeah. Let's go for it. Ectrimble, it's ultimately your call, as it's your code.

2/5 so far. Just need the 3 modders...

Sounds good to me, you might want to throw Frost's optimization their way too.

My optimizations is not ready for prime time, plus I haven't touched the server-side version yet.
This is ectrimble's domain. See what he wants to do.

First, who's Lordhood.

Secondly, I've only started to develop the server-to-server stuff, all the monitor stuff is in place (except the global ban stuff, that's been a real PITA).

I'd be ok with working along side them to develop this as I could definitely use more servers for testing. The basic layout of what I need is in place, I just haven't had the time to jump around between my computers and try to make them talk.

Ultimately, I'd like to know the who's and what's of their dev group and process and then go from there.

Then it's settled. Someone call up Alblaka.

Doc, was there anything else regarding the topic at hand that I need to know about? Do I need to get in touch with anyone about this or will this play out at another persons leisure?

And to the OP, don't hijack this thread, it has information I would like to know about and if it get's cluttered with crap about some other Mod, I'll just start nuking posts.

Shivedit: No need.

I'm Lordhood. I spend a lot of time on Industrial Rage nowadays, so I also frequent the server's forum. When I saw xKillerbees' idea, it reminded me of what you are trying to accomplish with Futurecraft. My thought was that this was a big chance for Futurecraft to reach a major milestone, not to mention boost the mod's publicity. Of course, I'm not expecting everything to happen now. I believe that, until everyone is completely comfortable with the server-to-server stuff, the best that can be done is letting xKillerbees know that you can make this idea become reality.

As for the global ban stuff, IR is part of an organization of IC2 servers with a universal ban list, so I think that once the additional servers are up, they'll also go on the list. Should make things a heck of a lot easier there. If you want to know more about Killer's dev team, I think he could probably shed more light on that than I can. If you want to take a look at the OP on the IR forum, here's the link: http://www.killerbeesgaming.com/forum/m/2959419/viewthread/2034606-server-fix-idea

Oh, and if anyone needs any motivation for doing this, look here:

Spoiler:

Well, I read through Killerbee's post, seems that he's doing something quite a bit different then what we're actually trying to do.

He's trying to do clone/separated worlds linked together in an innovated, but quite redundant approach to fixing Mojongs inability to properly implement multi-threading. Redundant because it's an overly complicated fix that technically doesn't fix anything, it just spreads the load out.

Now, I've made progress with connecting two servers together, you sort of have to build your own connector/listener pieces as the stock packets are pretty much useless, build yourself a custom packet, pass the information you need, then rebuild it on the other side. It's a lot of code and a lot of over head, although I've gotten it down quite a bit by splitting it off into multiple threads and making the connection and listener use their own threads so it doesn't bog down the main MC thread.

In reality, the problem won't be so easily fixed. My biggest issue right now is that the amount of hacking and slashing into Mojong's code that is necessary for all our pieces to come together will make porting it to a new version a tedious process, quite a bit more tedious than a normal mod loader or forge mod that just has to implement any new methods and they're good to go. Hence why I'm still working with 1.0.0 and haven't bothered moving to 1.1, I don't enjoy rewriting code because Mojong can't figure out how to make something work and every time I seem to fix an issue, Oh lookie here! A new version of MC.

So I suppose in conclusion, I'm quite open to the idea, it's a fun challenge to me, though I think it won't truely solve the problem as what happens when all 3 of your linked servers have 300 people on them? Opps, problems back. Anyway, I suppose I'll look into it and possibly drop KB (that's his new nickname in my posts from now on ) a line and see what exactly we're dealing with.

Okay, now I went and read the post. Definitely the wrong way to go about it, though I think his mindset was how to do as much as possible with the least overhauling. It doesn't even seem that hard. You figure mob spawning, pathfinding, plant ticks, and most tile entities are asynchronous. Chunk generation is completely asynchronous by design, and with not too much work you could split up redstone and IC2 components with pretty basic muxing. And then there is the netcode itself, but I rant about that often enough. It always surprises me how shoddy the back-end code is in the performance department, and more than that, how little attention it gets from the developers.

I'm facing similar problems to you at the moment. It's hard to make progress without resorting to hacks. Makes me want to tear out everything and start from scratch, but I couldn't do that either without a fuller understanding of the engine. Sucks, so I guess I can sort of see why Mojang doesn't want to deal with it, but come one guys! I figured out how optifine worked in like a day, what's the holdup? Fix your stupid game!

I read the post and I agree with Ectrimble and Fr0stbyte. What he's doing is linking multiple servers running on different cores on the same computer, not linking multiple servers over the internet.