Anticheat Performance Metrics

Jul 13th, 2021 (edited)
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 2.49 KB | None | 0 0
  1. NOTE: This test is run on an OVH vps. If you plan on running a server with more than 50 players, you should consider using a CPU that was made in the last 5 years. It would be unfair to actually run a server with more than 50 players on the potato that is my VPS. Take these numbers as more of a comparison between projects than a real world cpu runtime that would occur on your server.
  3. Test Parameters:
  4. - 50 OQ.minebots set to wander within a 25 blocks radius in an empty glass cage
  5. - Disregard netty channel standalone write() locking numbers as this can inflate the results.
  6. - Disregard any timings from dependency plugins as any use of them will be shown in the same results as the anticheat standalone in the spark viewer.
  7. - How test is run: Anticheat and any depends are loaded, bots are all logged on, bots teleported to middle of glass box, bot wandering is turned off, after 20 seconds of wandering, profiler is run for 90 seconds.
  8. - Anything that runs a thread sleep function on its own thread was removed from results. As long as these thread sleeps are isolated on their own threads, they will not effect anything.
  9. - Timings from depends like ProtocolLib and Atlas are not included as they include duplicate results from the anticheats themselves.
  10. - Packetevents functions are filtered from any anticheats that have packetevents built into the anticheat itself without external depend. This was a bit harder to do so anything that just uses packetevents could be up to 3000ms faster (at most) than what this listing shows.
  13. Anticheats are ordered by lightest to heaviest in total CPU runtime within 90 seconds (90000ms) with 50 players:
  15. NoCheatPlus: (5452ms) [This seems abnormally light, but this is the only anticheat tested that does not use obfuscation. Kauri actually tested 6600ms consistently in the same testing parameters but without obfuscation applied]
  16. Kauri (2.14.0): (9720ms total)
  17. Alice (1.2.2): (10224ms)
  18. AAC (v5.0.14): (15968ms) [v5.1.0 would not load for some reason]
  19. Spartan (Build 418): (42376ms)
  20. Vulcan (2.2.0-HOTFIX): (75464ms)
  21. GodsEye (Build 47): (145444ms) [This actually comes from Godseye really liking to abuse the hell out of SimplePluginManager's plugin api and Bukkit event system and its kind of sad because it would be pretty light otherwise]
Add Comment
Please, Sign In to add comment