You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
/// Object Pool for networked objects, used for controlling how objects are spawned by Netcode. Netcode by default will allocate new memory when spawning new
12
-
/// objects. With this Networked Pool, we're using custom spawning to reuse objects.
11
+
/// Object Pool for networked objects, used for controlling how objects are spawned by Netcode. Netcode by default
12
+
/// will allocate new memory when spawning new objects. With this Networked Pool, we're using the ObjectPool to
13
+
/// reuse objects.
13
14
/// Boss Room uses this for projectiles. In theory it should use this for imps too, but we wanted to show vanilla spawning vs pooled spawning.
14
-
/// Hooks to NetworkManager's prefab handler to intercept object spawning and do custom actions
15
+
/// Hooks to NetworkManager's prefab handler to intercept object spawning and do custom actions.
Copy file name to clipboardExpand all lines: CHANGELOG.md
+1Lines changed: 1 addition & 0 deletions
Original file line number
Diff line number
Diff line change
@@ -11,6 +11,7 @@ Additional documentation and release notes are available at [Multiplayer Documen
11
11
### Changed
12
12
* Replaced our polling for lobby updates with a subscription to the new Websocket based LobbyEvents (#805). This saves up a significant amount of bandwidth usage to and from the service, since updates are infrequent in this game. Now clients and hosts only use up bandwidth on the Lobby service when it is needed. With polling, we used to send a GET request per client once every 2s. The responses were between ~550 bytes and 900 bytes, so if we suppose an average of 725 bytes and 100 000 concurrent users (CCU), this amounted to around 725B * 30 calls per minute * 100 000 CCU = 2.175 GB per minute. Scaling this to a month would get us 93.96 TB per month. In our case, since the only changes to the lobbies happen when a user connects or disconnects, most of that data was not necessary and can be saved to reduce bandwidth usage. Since the cost of using the Lobby service depends on bandwidth usage, this would also save money on an actual game.
13
13
* Simplified reconnection flow by offloading responsibility to ConnectionMethod (#804). Now the ClientReconnectingState uses the ConnectionMethod it is configured with to handle setting up reconnection (i.e. reconnecting to the Lobby before trying to reconnect to the Relay server if it is using Relay and Lobby). It can now also fail early and stop retrying if the lobby doesn't exist anymore.
14
+
* Replaced our custom pool implementation using queues with ObjectPool (#824)
14
15
15
16
### Cleanup
16
17
* Clarified a TODO comment inside ClientCharacter, detailing how anticipation should only be executed on owning client players (#786)
0 commit comments