Minecraft 1.8.1 Performance / Benchmarking Tests of 1.8.1 and a bit of 1.7.3! =(

Discussion in 'Bukkit Discussion' started by Kane, Sep 30, 2011.

Thread Status:
Not open for further replies.
  1. Offline

    Kane

    Not sure if this is really truly much of the best best benchmarking I guess it works I do take the steps and repeat them exactly each time. I start from spawn looking the exact same direction (west) Then after things have loaded and settled I run jconsole and I walk west 1000 blocks. Then I idle for a few seconds and log off. Then I idle a few more then login. Then I idle till most the chunks load and log off once more. You can easily see the log off as the 2 last spikes u sally hitting 2% cpu is the 2 log offs.

    Also this pre generated Map from Godcraft so everything every time is the same conditions even the time.

    For the test I had disabled a few plugins that use mysql but I have in the past and present tested their performance and they seam to have not much effect besides logblock and even it seams to not cause much slowdown these days.

    I will list the plugins under the image to give you an idea of what I was using and its version at the time.

    I also want to add the specs of this test server were a 2600k with 16gigs of ram running on a Gen 3 SSD. Their was no fooling around and this was a private test lan server no other connections.

    Please note there was no spawns unless stated near the bottom.

    TONS MORE INFORMATION AT THE NEXT POST!

    Godcraft Using Bukkit 1.8.1 1185 + Spout with Smooth Teleporting On:
    [​IMG]

     
  2. Offline

    THG3

    This is a very informative post.
     
  3. Offline

    Revelator

    Cpu load seems better tho, instead of 20, it was up to 20 _once_, mem load was quite ridiculous tho, everything from 2 to 11GB, but yeah have noticed the chunk load issues.
     
  4. Offline

    gameswereus

    Thanks for this post!
     
  5. Thanks for the fantastic input!, @Kane !
     
  6. Offline

    Rwembee

    @Kane what are the command line options and java version that you use? Just curious.
     
  7. Offline

    Nathan C

    Is this with 0 players on?

    NVM, I see, nice!
     
  8. Offline

    desmin88

    Great post, very informative.
     
  9. Offline

    Theborg

    Neat to get the facts down, sadly enough ever since the first dam version of the Minecraft in this form.
    Minecraft Beta has been taking a liniar path down performance road, and not expecting a change in that anytime soon.
     
  10. Offline

    sukosevato

    The more features they add the more stuff gets added that listens on the main server thread -> the worse performance gets. I'm hoping that notch will use the month of code freeze to move a lot of stuff to seperate threads and to optimize the code a lot.
     
  11. Offline

    Theborg

    Someday maybe, in the meanwhile we need to wait for custom servers.
     
  12. Offline

    davr

    Man you are gonna cry when 1.9 comes out...I'm running 1.9 prerelease 3 vanilla server, and with zero people logged in, the server uses 50-100% CPU (of one core, quad core i5-2400 3.1 GHz server running linux). I don't think I've ever seen it drop below 50% cpu no matter what.
     
  13. Offline

    sukosevato

    Not really I think, Bukkit is already a custom server. I think if we really want to see performance going up we need to get a lot more mulithreading. Because personally I think a game like minecraft will never be finished.

    If the game would ever be finished it might be worth it porting it to C or C# or something. But if stuff keeps getting added you don't want a language like that since it'll just say fu and won't work if there is a bug. While with Java its much more forgiving.

    So I think our best bet is just getting it as multithreaded as possible and between every 'wave' of new features force our selves to do a couple of months of 'feature freeze' in which we work on optimizations.
     
  14. Offline

    Theborg


    Your wrong in that, Bukkit is a system build off the original minecraft server. Bukkit isn't at fault, its the main server system that its all build on. A custom server would need to rewrite the original vanilla code.
     
  15. Offline

    Greylocke

    If chunk loading is the culprit, I wonder if the caching features of the Spoutcraft client will help the situation?

    Secondly-- I'm running a similar set of plugins to what @Kane listed. But I take terrible hits to my CPU. To the tune of about 8-12% CPU usage per user on a 2.4GHz core. This seems insane. I've tried pulling out what I thought would be heavy-hitters, but only see marginal differences.
     
  16. Offline

    NotYetRated

    Awesome post Kane! Treacherous path we are headed down in the performance department haha....

    When you "Turned off all drops, remove all drops" What did you use to do that exactly?

    Also, I use that lagmem program too. I have found it useful to see in graph form, as it gives you TPS/MEM/PLAYERS all in one... Funny to see my RAM usage continuously go up even as player count holds rather steady.... Memory leaks I am assuming... Awaiting time to investigate WHO is causing it though.

    My Plugins:
    Code:
    Questioner    : 0.4
    FalseBookCore : 0.88alpha
    WorldGuard    : 5.3
    HeroChat      : 4.10.3
    KitPlugin     : 2.0
    FalseBookBlock: 0.88alpha
    WeatherControl: 1.0
    CommandBook   : 1.6-customized
    iConomy       : 5.01
    LagMeter      : 0.5
    Spout         : 1.0.6.403
    SimpleBan     : 2.1
    WorldEdit     : 4.7
    FalseBookIC   : 0.88alpha
    Vault         : 1.0.0dev-b41
    Minequery     : 1.5
    NoCheat       : 2.08c
    PermissionsEx : 1.14
    xAuth         : 2.0b4.3
    AutoMessage   : 1.5
    SpoutMOTD     : 1.4.3
    Permissions   : 2.7.7
    ecoCreature   : 0.0.6c (Mo' Mobs!)
    dynmap        : 0.23
    LWC           : 3.52
    ChestShop     : 3.12
    MyHome        : 2.0.1
    Register      : 1.5
    SpoutBackpack : 2.3.5
    WorldBorder   : 1.4.3
    HawkEye       : 1.0.5
    NoLagg        : 1.26
    MobArena      : 0.94.3.11
    Jail          : 2.0.1
    Towny         : 0.76.2.0
     
  17. Offline

    Kane

    World guard offers a wonderful option stoplag or lagstop forget what one and it will kill so many things. then add -c at the end to start it up again.


    But the end of the day 99% of TPS dropping is chunks and nothing else.
     
  18. Offline

    NotYetRated

    I have noticed the chunks are the primary cause of me losing TPS also, was just curious as to the removing of all drops. Thank you! Awesome post.

    Edit: I had originally hoped clearing drops would clear up my RAM usage/leak issues but alas it did not.
     
  19. Offline

    Rwembee

    I've been wondering if it is the new lighting and or biome code. The server seems to do the lighting rather than let the client, at least the world pregen tools lead me to believe.
     
  20. Offline

    NotYetRated

    Actually, looking at your lagmem log in more detail, I also see you have the similar RAM situation as me. Where RAM usage will fluctuate wildly even with relatively constant members online. Perhaps this is normal? Although I suppose "online members" means little when their location and how close together/far apart each member is would be a driving factor in memory usage.
     
Thread Status:
Not open for further replies.

Share This Page