Is it possible to use this system generation with Mirror multiplayer system? about marching-cubes-on-unity-3d

[Javier-Garzo]

I create a diagram of the network idea of edit replication between all clients:

[Javier-Garzo]

thanks this is really good method but if i only send actions instead of chunk data, there might be a problem caused by synchronization. But if i remember correctly minecraft is struggling with that problem too. If you have a lag then someone place block before you. But since its a server auth environment, client has to obey server so there wont be desyn i hope

Yeah the idea is use the host as authory and for merge some edits (marching cubes world modification support multiple actions in the same point). So the system support better the lag problems that Minecraft world.

[Javier-Garzo]

The current system not have a multiplayer implemetnation but it's possible to support one. The main point is send three elements over the network:

1. The world configuration: it's the class inside the NoiseManager used for the creation of the new chunks (not edited ones). You need send this one time, when the user connect to the server.
2. The ChunkData: The chunk data contains the terrain data (all the modifications applied to the the terrain), in this case the server must send that data when a connected client load that area of the world. The data is in : Application.persistentDataPath >worlds > [WorldName].
3. The real time modifications: You need replicate all the modifications from a client in the host and the other clients for the same state of the terrain in all users. Function that modify the terrain: chunkManager.ModifyChunkData(). You need send this each time user make a modification in any area of the game or to the current areas of the client and send the chunk data when client render a modified chunk.

The implementation is possible however you need a correct knowledge of the code of the project..

[Diamantino-Op]

Thx, for the answer, i'll try to implement those :)

[Javier-Garzo]

Any more doubts about this you can create new issue or continue in this one. Try to implement the three elements in the order of the list in my previous comment I think that's the easier order.

[emre-sahinn]

Hi @Javier-Garzo , I've implemented multiplayer chunk system but chunk's data is too much for MTU. I tried to compress chunk data by using LZ4 and it compressed chunk data to 1k byte. But sometimes compression not works well and exceeds MTU which is roughly 1024 byte.
How can I compress chunk data better than generic compression algorithms?

Also I tried to implement lighting like minecraft, light spreads out to chunk vertices etc but the problem is: When player modifies chunk, we need to calculate light each time(using unity jobs). Do you have any idea about how can I improve performance?

[Diamantino-Op]

i'm working to get it working with Mirror

rn it can succesfully load chunks and work like normal on Host mode

(Doesn't need to send files with Mirror)

[emre-sahinn]

@Diamantino-Op if your world is not modifiable then its easy to make it multiplayer but my world is editable and players can dig/build.

[Diamantino-Op]

mine too

[Diamantino-Op]

basically it loads world like normal, it spawn the chunks only on the server btw, and the server sends them to the client in real time

[emre-sahinn]

basically it loads world like normal, it spawn the chunks only on the server btw, and the server sends them to the client in real time

did you check the chunk data size :D
its too much to send over network mine was 70k byte which is 70x bigger than MTU

[Diamantino-Op]

nope didn't check

[Diamantino-Op]

i think you can set the max message size here

[emre-sahinn]

but you are using TCP based transport which automatically splits one big packet into chunks. Problem is not about sending chunk data, problem is its size is too much to send over network. Image you have 100ccu and players are wandering around the world. Its nearly impossible without some compression.

[Diamantino-Op]

then how can we do it?

[Javier-Garzo]

I don't know if it's possible to get a better comparison of the chunk I used in the CompressHelper.cs the Gzip algorithm, but I think LZ4 is better.
Similarly for the illumination I think you need recalculate the illumination each time because you build a new mesh each modification.

For the network chunk size problem, maybe I have a solution:
The idea is that you send to the user all the region data when he connect to the server, so the user have the current state of the game when end the load. At that point all edits action are sent to the server and this one use it to update his chunk data and send a compressed version of all edit actions in that frame to all clients. In this system the client must have all the region data loaded (or all regions where a user is playing) so he can modify the data of that chunk when the server send the modifications (low cost part). The client only creates the mesh of the user active chunks (that's the high cost part).

[Diamantino-Op]

i have done this for now if you are ineterested: https://github.com/Diamantino-Op/Mirror-Multiplayer-Macrhing-Cubes-Terrain

[Diamantino-Op]

this is my version (Has some bugs but "works")

[Diamantino-Op]

8Kbs network usage when loading a lot of chunks together

[emre-sahinn]

I don't know if it's possible to get a better comparison of the chunk I used in the CompressHelper.cs the Gzip algorithm, but I think LZ4 is better. Similarly for the illumination I think you need recalculate the illumination each time because you build a new mesh each modification.

For the network chunk size problem, maybe I have a solution: The idea is that you send to the user all the region data when he connect to the server, so the user have the current state of the game when end the load. At that point all edits action are sent to the server and this one use it to update his chunk data and send a compressed version of all edit actions in that frame to all clients. In this system the client must have all the region data loaded (or all regions where a user is playing) so he can modify the data of that chunk when the server send the modifications (low cost part). The client only creates the mesh of the user active chunks (that's the high cost part).

thanks this is really good method but if i only send actions instead of chunk data, there might be a problem caused by synchronization. But if i remember correctly minecraft is struggling with that problem too. If you have a lag then someone place block before you. But since its a server auth environment, client has to obey server so there wont be desyn i hope

[Diamantino-Op]

emre, try my code, it follow a bit the solution given by Javier

[emre-sahinn]

emre, try my code, it follow a bit the solution given by Javier

gonna check thanks

[Javier-Garzo]

8Kbs network usage when loading a lot of chunks together

The problem here is that is too much data trafic if you want scale the multiplayer. You only send the edits after the user end loading the region data on the initial connrection?.

[Diamantino-Op]

8Kbs network usage when loading a lot of chunks together

The problem here is that is too much data trafic if you want scale the multiplayer. You only send the edits after the user end loading the region data on the initial connrection?.

basically yes, i made mirror check if client has fully loaded the terrain

[Javier-Garzo]

Now the system allow the user edit the terrain each frame, maybe edits with a littel of cooldown between them reduce the traffic, for example wait the dig animation of the shovel end (1 edit per second).

[Diamantino-Op]

yeah, i may do that in the future