Hello Now we getting somewhere. True I don't now enough about this and thats why I keeping send messages here. There is really not too much to argue, after your answer :) I'll optimize that file cache and figure out to myself how this memory management works actually. But still I can't understand why 20/20 server isn't affected anyway and 28/28 server are ?
Snagu ----- Original Message ----- From: "Clayton Macleod" <[EMAIL PROTECTED]> To: <[email protected]> Sent: Wednesday, January 19, 2005 1:38 AM Subject: Re: [hlds] Windows 2003 server lag issue > Sorry, but you don't understand enough about memory management to > understand why your arguments don't hold up. A user process such as > HLDS has absolutely no idea about how it is using the disk cache or > the CPU's L1/L2 memory cache. In fact the OS doesn't even know what's > in the L1/L2 cache. A user process only asks for memory, and the OS > provides it. The user process doesn't know where it is in actual > physical memory address space. Only the OS's memory management knows > this information, because this information is irrelevant to the user > process. The user process gets its own virtual address space and > that's all it deals with. The OS's memory management does the handling > of this virtual address space and decides what goes into real memory > or gets shoved into the pagefile/swapfile. The user process has > nothing to do with it. And one user process can't affect another user > process. It doesn't matter what's going wrong with HLDS, there is > nothing another program can do to correct it. It can only be resolved > by fixing either the HLDS configuration or the HLDS programming > itself. Even the user process itself has pretty much nothing to do > with whether or not its data is in physical ram. It's highly unlikely > that a memory leak is causing any performance problems. Memory leaks > don't really affect anything except memory usage. If you have a memory > leak it just means that you have more virtual address space allocated > than you are actually currently using. It doesn't necessarily mean > that you are using more ram. Just means you are using more virtual > address space. Since it is a memory leak we're talking about, the > memory pages won't be 'touched' and since they aren't being touched > they will be candidates for moving to the pagefile/swapfile and out of > ram very quickly. And once they're there they aren't doing any harm to > any other processes because they're no longer using up physical ram. A > memory leak doesn't make anything slow down other than the short > period of time it takes to actually write them out to disk. It's not > like a memory leak is going to make you drop 50fps. It doesn't cost > you any CPU. It's just extra memory address space that wasn't > released, almost meaningless, except if it continues leaking and you > eventually run out of pagefile/swapfile space and after that run out > of actual ram. > > You might as well delete your memory management utility, because it > isn't helping you do anything. The only thing it could possibly do is > adjust the size of your disk cache, and even that isn't helpful. The > OS already automatically adjusts disk cache size depending on demand. > If user processes need the ram more than the disk cache does, then the > user processes get that ram. If the user processes aren't using all > the ram then it gets assigned to the disk cache where it can benefit > the machine, rather than sitting idle and being wasted. More and more > people are writing memory management utilities like those, and more > and more people are using them. Even though they don't do anything but > hurt your overall performance. But, that's what happens when people > don't understand the technology at hand, someone else can make them > believe something because they don't have the knowledge at hand to > know better. > > > On Tue, 18 Jan 2005 23:32:29 +0200, Snagu <[EMAIL PROTECTED]> wrote: > > Nice, > > > > But I have to disagree with you. True that memory management software > > can't affect HLDS-process or memory it's using at the moment, but it could > > optimize memory cache / CPU L2 cache usage etc., so if there is some > > problems > > how HLDS uses or releases these resources, then this will affect how > > gameserver works. > > I haven't talk anything about performance problem, I don't have it, there is > > something else > > going on. Anyway I have spend all day and watching how server works. First > > day I haven't > > had to reboot hlds-process, because lag ( after installing that memory > > manager software ). > > You seem to be so shure, then maybe you have the answer ? > > > > Not blaming, just searching... so no need to be so aggressive, peace man :) > > > > Snagu > > > > ----- Original Message ----- > > From: "Clayton Macleod" <[EMAIL PROTECTED]> > > To: <[email protected]> > > Sent: Tuesday, January 18, 2005 10:56 PM > > Subject: Re: [hlds] Windows 2003 server lag issue > > > > > except for the fact that no software you install can do anything > > > directly with the hlds software or its memory, so whatever you've > > > installed isn't likely to help in any way whatsoever. The only thing > > > that can deal with memory directly is windows' own memory manager, and > > > the only thing that can deal with any given program's memory is only > > > that program itself. The only thing any add-on 'memory management' > > > software can do is for paging to happen earlier than it otherwise > > > might happen, and even that isn't going to affect game server > > > performance in any way whatsoever. > > > > > > > > > On Tue, 18 Jan 2005 22:10:06 +0200, Snagu <[EMAIL PROTECTED]> wrote: > > > > Yep,yep > > > > > > > > Reducing number of players or changing timer values > > > > didn't help a bit. I started to measure performance counters > > > > and there seems to be soft faults rising (memory / transition faults) > > > > same speed as lag is happening. So there seems to be memory leak or > > > > some other memory related problem with hlds-process. I installed > > > > memory manager software and managed almost get the hole lagging problem > > > > away. > > > > Still there is very little problems, which disappears after process > > reboot. > > > > I understand that DoD is not number one in interest list of Valve and > > source > > > > is coming, but > > > > still maybe some advices or explanations related this problem from Valve > > > > would be nice ... > > > > > > > > Tired to chase ghosts... > > > > > > > > Snagu > > > > > > -- > > > Clayton Macleod > > > > > > _______________________________________________ > > > To unsubscribe, edit your list preferences, or view the list archives, > > please visit: > > > http://list.valvesoftware.com/mailman/listinfo/hlds > > > > _______________________________________________ > > To unsubscribe, edit your list preferences, or view the list archives, please visit: > > http://list.valvesoftware.com/mailman/listinfo/hlds > > > > > -- > Clayton Macleod > > _______________________________________________ > To unsubscribe, edit your list preferences, or view the list archives, please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
