TunerZ (a project in development)

[Angelo]

Hello!

It's a shame that PGD (the best forum) is dieing.
Aside that I'm proud to pronounce my new project, "TunerZ".

The game will be a variation of Racing, Simulation, Tuning and RPG.

Currently all the attention is going to the documentation which can be downloaded here (still in progress as you might have noticed).

The project will be a multiplayer game, with all the race-factors within it.
TunerZ will use the Network Library written by myself.

The server is currently in the following state:
- multithreaded
- can handle 1471 online players (when there are 1472 players online the server will crash)
- can handle up to 200 players that connect to the server at once (this is at my pc, my computer runs out of memory)
- saves/loads data from a SQL database Server

The client (the game) is currently in the following state:
- can connect to the server
- can login

What needs to be done within a short period:
- Carphysics (50% done so far)
- Optimize the Network Library

What needs to be archieved within a short period:
- Server must be able to handle around 10.000 online players
- Server must be able to handle up to 1.000 players that connects at once.*

* = Archieved yet.

Further information will be shown.

Angelo.

[WILL]

It's a shame that PGD (the best forum) is dieing.

:lol: How did you determine this? The site goes through quiet periods, it's nothing new.


So is your game going to be focused more on realism or on gameplay?

[Angelo]

Yes, whatever , well I noticed that my post of 2 days ago was still in the recent topics
And it seems to happen more and more, so my conclusion is dieing

So is your game going to be focused more on realism or on gameplay?

There will be a balance between these 2 elements.
I think the realism shouldn't suppress the gameplay.
This balance can only be found by testing and testing and testing...

[savage]

What SQL database are you using?

[Angelo]

Microsoft, I think it's the most userfriendly and easiest database.

[Angelo]

I got new results from my server, fresh results.
And I must say, I'm very satisfied with it!!!

The previous results (1471 players online) were made 1 week ago.

Today I'm glad to announce my new results:

- Maximum players online: 6837
- Maximum players that can connect at once: 623

These results are archieved by the same pc, under same circumstances.
There are a few things I've changed:

• - The interval for every check-thread: the more users, the higher this interval should be. By doing this there will be lag, but the server will be able to handle more players. I've made it variable from 0ms to 100ms. These results are archieved by a interval of 100ms.
  - The priority of the server is set to highest, instead of normal.

So with a variable thread interval the server is capable to handle more players.

EDIT:

After a retry (the server just finished the simulation) there were 7016 players online, then the server started refusing connections.
The maximum players that were able to connect was 652.

The improvements are coming!

[Angelo]

After a long night tweaking my server, I've been able to archieve one of my goals.
The server can handle 1.000+ connections at once now.

But I'm wondering, at the moment my server is multithreaded.
When a player connects, a thread is been created.
This means, 1.000 players = 1.000 threads.
In my opinion this is overkill, isn't it much better to create a singlethreaded server?
Please let me know all your opinions...

[technomage]

Modern operating systems can handle lots and lots of threads, Webservers use multithreading so I don't see it being a problem for your system.

[Angelo]

Okay, thanks for your reply.

But, wouldn't the multithreading suppress the speed of the server?

[JSoftware]

I guess you are using windows in which case your only realistic option would be to use threads. I don't think that asynchronous operation would be realistic on a scale like that.

And blocking seems to me like the root of all evils..

1000 threads is nothing. I've tried to have a program running which would spawn 25000 freerunning threads at once. Granted, I didn't use schemes like synchronization or the like, but it ran without any problems