app.publish is faster than ws.send?
00sos00 opened this issue · 11 comments
I am currently working on a game, and I am sending the info of all 100 players to each other.
The way I am currently doing this is by doing
for (ws in wsArr) {
ws.send(data, true); // data is around 2000 bytes big
}I benchmarked this loop and found that it takes around 12ms.
But when I switched to this
app.publish('global', data, true) // same size of dataIt only took around 3ms
So what's the catch here?
Why would there be a catch?
Why would there be a catch?
Meant to say what's the reason behind this
It's the same question just worded differently ;)
Pub/sub is faster because it's designed to be faster for Pub/sub. It's a misconception that it would be the same as a for loop.
Shitty libraries implement it as just a for loop. But you probably want to benchmark end to end.
The websocket benchmark of https://bun.sh uses Pub/sub thats why it performs way better
benchmark
Interesting, does pub/sub use .send under the hood? And if so, would it be possible for me to design a similar architecture but with a custom message for each client?
Typically these "IO games" send the same message to everyone but the sender, for every client.
Typically these "IO games" send the same message to everyone but the sender, for every client.
Depends on the game. Some games send only entities that are within a certain distance from each player
It's the same. You just divide the game world into squares and have those be their own topic. Still overall same idea
It's the same. You just divide the game world into squares and have those be their own topic. Still overall same idea
Oh that's clever, never thought of splitting world just for sending messages. I always thought of it as a way to reduce collision checks
Is there a limit to how many topics I could create? I am planning having a 21x21 grid which is around 400 topics
You can have many many topics but always good to benchmark