Server Performance

2»

Comments

  • MOOtantMOOtant Join Date: 2010-06-25 Member: 72158Members
    <!--quoteo(post=1859976:date=Jul 12 2011, 10:20 AM:name=endar)--><div class='quotetop'>QUOTE (endar @ Jul 12 2011, 10:20 AM) <a href="index.php?act=findpost&pid=1859976"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->Interesting, me too. Have you tried not binding(for lack of a better word) to a single core? I know that the server does have some multithreading capability, so you would be missing out on that by binding. Here is what I have currently with 18 players.
    <a href="http://imageshack.us/photo/my-images/809/cpuje.png/" target="_blank"><img src="http://img809.imageshack.us/img809/4186/cpuje.png" border="0" class="linked-image" /></a>

    What about windows Processor scheduling? I believe default for server editions is "Background Processes" which is what I currently use. You?<!--QuoteEnd--></div><!--QuoteEEnd-->
    The fact that you see use on several cores doesn't mean it uses several at once. It might be just cycled between them all the time and giving impression of speedup. No matter what you'll do with Windows it won't make server use several threads to do useful work (I'm not counting stuff like network send/receive thread).

    I have it set to background processes.

    maesse's profiling shows that TraceRay and family are simply badly written. 9000 traces for MACs per tick is quite absurd too.

    Whose idea was it to enable console only after '`'? I had trouble with displaying anything at all (server DirectX+RDP) and when server.exe finally showed up that window didn't give slightest hint that it waits for key press. Why would you start with -console, have D3D window running and not want to see the console?
  • FlayraFlayra Game Director, Unknown Worlds Entertainment San Francisco Join Date: 2002-01-22 Member: 3Super Administrators, NS2 Developer, Subnautica Developer
    We're definitely still working on server performance in general. Brian was working on something yesterday that we think will help (updating minimap blips, of all things). But we're going to keep working on this until it's running great.
  • falcfalc Join Date: 2011-03-18 Member: 87128Members
    A seperate server forum would be great though! ;)
  • endarendar Join Date: 2010-07-27 Member: 73256Members, Squad Five Blue
    <!--quoteo(post=1861607:date=Jul 20 2011, 12:41 AM:name=MOOtant)--><div class='quotetop'>QUOTE (MOOtant @ Jul 20 2011, 12:41 AM) <a href="index.php?act=findpost&pid=1861607"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->The fact that you see use on several cores doesn't mean it uses several at once. It might be just cycled between them all the time and giving impression of speedup. No matter what you'll do with Windows it won't make server use several threads to do useful work (I'm not counting stuff like network send/receive thread).

    I have it set to background processes.<!--QuoteEnd--></div><!--QuoteEEnd-->
    Actually i kind of meant the opposite. I see no swapping between cores, the image shows almost exclusive use on core 0.

    By binding directly to a core, you will stop it swapping between them, but you are also limiting everything to one core. According to Max there is the ability for some processing on other cores. In response to the "Asynchronous Physics" that went in around build 162:
    <a href="http://www.unknownworlds.com/forums/index.php?showtopic=110837&view=findpost&p=1822558" target="_blank">"It means that the physics happens in parallel with the rest of the game and can use a separate processor if available."</a>
    Although I'm not sure if that refers to the server or client, or both.

    And I agree with falc, makes it easier to keep track of these posts instead of having to dig them up each time.
  • MOOtantMOOtant Join Date: 2010-06-25 Member: 72158Members
    I never saw more than 23-24% use when it was free to use separate cores. It used the most (25%) when set to 1 core. And no I don't exactly care about some comment to some question on forums 20 builds ago. I tried affinity set to 1 core, 2 cores and so on. Unless Windows scheduler screws up even such simple scenario as 1 process, 2 busy threads, I'm sure that NS2 doesn't get measurable speedup from 2+ cores.

    Maybe the server I am using has other processes that cause task scheduler to swap between cores.
  • A[L]CA[L]C Join Date: 2010-07-25 Member: 72801Members
    We just had a big match which lasted ages. The server was lagging extremely hard... until we killed the arcs. Once the last arc was gone, the lag went too.
  • RuntehRunteh Join Date: 2010-06-26 Member: 72163Members, Reinforced - Shadow
    Varies depending on server.
  • A[L]CA[L]C Join Date: 2010-07-25 Member: 72801Members
    Not sure if this would be causing it, but the arc was stuck in the wall and unable to move.
  • McGlaspieMcGlaspie www.team156.com Join Date: 2010-07-26 Member: 73044Members, Super Administrators, Forum Admins, NS2 Developer, NS2 Playtester, Squad Five Blue, Squad Five Silver, Squad Five Gold, Reinforced - Onos, WC 2013 - Gold, Subnautica Playtester
    Considering how the pathing currently works, a stuck arc could cause a spike. How an arc performs targeting could also be a culprit.
  • MOOtantMOOtant Join Date: 2010-06-25 Member: 72158Members
    edited July 2011
    When I started dedicated server with console and profiling on I got this:
    ExitNode for EffectManager:InternalGetEffectMatches doesn't have matching EnterNode
    and whole server was spiking down to 0.1 ticks and messages/s. Disabling profiler "fixed" it.

    This dedicated server has some kind of Intel graphics card and I don't know if it supports anything at all.
  • wulf 21wulf 21 Join Date: 2011-05-03 Member: 96875Members
    <!--quoteo(post=1862026:date=Jul 21 2011, 12:36 AM:name=McGlaspie)--><div class='quotetop'>QUOTE (McGlaspie @ Jul 21 2011, 12:36 AM) <a href="index.php?act=findpost&pid=1862026"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->Considering how the pathing currently works, a stuck arc could cause a spike. How an arc performs targeting could also be a culprit.<!--QuoteEnd--></div><!--QuoteEEnd-->

    So maybe there is a bug that causes ARCs to constantly recalculate the path to the move target if no valid path has been found?
Sign In or Register to comment.