Peer-to-peer

This is an experimental feature, it is subject to change and may contain bugs. Your feedback is welcome to improve this feature.

GDevelop supports peer-to-peer (P2P) connections to enable basic multiplayer games. This works through the concept of remote events. Connect multiple instances of the game using their ID, before remotely triggering conditions on the other instances.

This extension is not suitable for all types of multiplayer games and is potentially unsafe against cheating. As every instance of the game is connected directly to each other, there is no machine that can be considered trusted more than anothor.

As such, it is hard to verify if someone is cheating (unlike dedicated servers). P2P supports a maximum of 250 simultaneous connections, and is therefore also not suited for very large multiplayer-focused games like MMOs.


Selecting a broker server

A running game, called a client, must connect to other clients before being able to send data. For this, it needs a way to self-identify and find other clients. To do so a broker server must be configured. It's a server with a fixed, well-known address that stores all the addresses of the connected clients and give them to each client, so that they can connect to each other.

There are two options for setting up a broker server:

  • Setting up a custom server (recommended), which can be run on a local computer as a test.
  • Using a default, public server.

Set up a custom (local) server

A local server can be set up easily. install Node.js will need to be installed. The LTS version is recommended.

Open a command line. To do so on Windows:

  1. Press the Windows and R keys. This will open the run window
  2. Type cmd in the popup.
  3. Press the enter key.

After opening a command line, complete the following:

  1. If this is the first time completing these steps, to install the server type npm install peer -g
  2. After the server is installed, and every subsequent time (to start the server), type peerjs -p <the port>
    • Any port that is not already used by the computer can be used.

Other options can be passed to the PeerJS server, see its website for more information.
After the above has been completed, the Use custom broker server action is used to connect to the server.

  • localhost can be used as the host address to point to the local server.
Note that this is a local server, so it will only work on your machine. When releasing your game you will need to deploy one to a Node.js compatible hosting, like Heroku.


Use the default server

It is not recommended to use that server, you should use your own if possible. The default server is not operated by GDevelop and GDevelop team is not responsible if anything goes wrong using that server.

You can also use the default server provided by PeerJS. To use that server use the action “Use the default server”.

Connecting

To connect instances, you need to enter their ID in the other instances. The ID can be found with the expression P2P::GetID(). To connect, use the “Connect to other instance” action and pass as parameter the ID of another instance. Both instances will then connect automatically. You can then send an event from one instance to the other one to make sure that the connection is established.

Interacting with connected games

Once you got connected, you can trigger actions remotely. You can select another specific game instance (using its id) or send an event to all connected instances.

Choosing if you want to activate data loss mode

You might be wondering what the “data loss” parameter is for. Due to how GDevelop is made, only one occurence of a remote event can be handled when the event sheet is executed (this happens roughly 60 times per second). To help optimize events execution, we provide the choice to use the dataloss mode.

  • With the no dataloss mode, every remote event is queued, and on every frame if there is one in the queue, we take the oldest one and handle it. This makes sure every data is processed/taken into account.
  • With the dataloss mode activated, it doesn't queue the data but only store the latest occurence of the remote event. This means only the latest data is processed and outdated data will be discarded.

Here are two examples:

  • if you use a synchronised score counter, you don't want to lose any data, as missing only one point of the counter would desynchronise the counters, so the dataloss mode would be deactivated.
  • If you want to synchronise positions, only the last position sent is relevant, not older positions. In this case, you would activate the dataloss mode to prevent delays/lags.