#Server performance improvement suggestions

1 messages · Page 1 of 1 (latest)

native wagon
#

My intention is to take a look at low-level updates and maybe implement some more direct optimizations than just using APIs provided, like moving parts of data management to native libraries, multithreading some processes, etc.

modest oak
#

While I don't really have any profiling results to share, from experience, in survival servers anyways, entities and chunk ticks cause the most lag.

You could technically simulate that by loading a bunch of chunks like when players are spread out, and then also spawning just a ton of entities. Not sure how much that will help you though, considering your hardware is more than likely different than the server's.

native wagon
#

I have some mid-performance servers available, I'm mostly lacking means to utilize them 🙂 Probably starting with entity tick is reasonable as I've experienced it myself, I'll see if I can move some of the logic to worker threads and somehow manage the data. Also, is it worth looking at collisions? I'd assume it's already grid-optimized in a block game, but maybe there are reasons 😉

#

Thanks Dessie for your input, appreciate it a lot 🙂

strong hawk
#

Are you talking about like in game changes at all? Or?

native wagon
#

Yeah, my intention is to fiddle with NMS directly and maybe squeeze something out of other cores

#

I'm not decided on anything just yet, just wanted to look around. I've heard hosting big monolithic server is an issue due to single-core nature of the server implementation, that's why we have bungee and the whole infrastructure around. I'm obviously not going for replacing those, some big minds put their time into that and I doubt that I'd beat that considering that I'm not very into minecraft modding and wouldn't even consider myself java a programmer xD My intention is to maybe offload some of the work done by game thread, possibly to native code (that's my area of expertise) and utilize multi-core benefits of modern server machines