Posts: 783
Threads: 12
Joined: Jan 2014
Thanks: 2
Given 72 thank(s) in 61 post(s)
The main reason we haven't got any data is that no one has yet produced a better way of load testing minecraft servers than just loading up a lot of clients which is extremely intensive work (you'll run out of compute resources for clients long before exercising a good server unless you have a lot of machines). So there aren't much in the way of benchmarks. Just semi anecdotal stuff like Howeners test. We'd love to see some stats if you could produce a test.
Posts: 4,624
Threads: 115
Joined: Dec 2011
Thanks: 693
Given 490 thank(s) in 421 post(s)
04-26-2015, 02:11 AM
(This post was last modified: 04-26-2015, 02:13 AM by NiLSPACE.)
MCServer can be a little unstable. Especially if you have entities or simulators enabled like redstone. Badly written plugin's can also cause a crash, so it's worth to test some plugins first.
I'm afraid you can't use the plugins and mods from Bukkit or Spigot. It's not only that the language is different, but especially the architecture. If you want a mod or plugin from Spigot/Bukit/Vanilla the author (or somebody else) has to completely rewrite it as either a Lua plugin, or integrate it in the server itself.
Posts: 783
Threads: 12
Joined: Jan 2014
Thanks: 2
Given 72 thank(s) in 61 post(s)
Is that 1,000,000 players in a single world or across multiple worlds through Bungeecord? Because if its one world you're going to need to modify whatever server you use to handle it. So that means working with Mojang, or using an open source server like MCServer. What are you trying to do with it and what are you're estimated server specs? Because that many clients is at the scale where you have to look at tuning the kernel in order to handle the data, or going to a cluster, and I don't know of any server that currently supports distributing a single world across multiple servers.