> Sorry to continue but there are two many issues left unanswered: > ----- Original Message ----- > From: "Chance Sullivan" <[EMAIL PROTECTED]> > > >> So why not just minimise them this would produce the same effect > >> according to your logic. > > It would not because doing it that way, your still using > the GDI and > > User resources where as if your running it as a service and > not using > > the Interact With Desktop option your not using them. > > Thats just a pile of rubbish. Start your server under a > service and query its resource handles with a debugger there > is no difference from using that run method to starting it > from the desktop you just cant see the output. If the server > was specificly written to run as a service yes this would be > the case but simple running it as one does not change is > resource requirements on single bit. I did query them, and the GDI and User Resources are at 0, otherwise I wouldn't have stated that they were.
> >> Using svrany will never leave memory allocated or unclosed > >> filehandles, dont know where u got this impression. > > That is incorrect as your not taking into consideration > applications > > that don't respond well to the Logon/Logoff events when > being ran as a service. > > Wrong! If a app doesnt ignore the WM_ENDSESSION message ( > which is what prompts them to close on logoff ) your app will > just close. > This won't leave any memory / filehandles lying around. Even > if an app crashes windows reclaims those. Not all of the time it doesn't as not all applications follow the proper procedures, or listen to the proper events, in a perfect world they would. Windows can and does have zombie processes occur like *NIX does whether it be from a foreground app, or a service, and using srvany has caused that to happen. > >> So you put 4Gb of ram in your game server machines, nice > but waistful > >> :P > > 2GB usually, sometimes 3 or 4 depends on the system setup and cpu's. > > Might be waistful to you, but If you a person that wants a memory > > buffer of about 256 to 500mb to account for spikes in > usage, sometimes > > more than 2GB is needed. > > Hmm 100Mb average per you said, 8 servers thats 800Mb you > still got 1.2Gb free on a 2Gb machine. Didn't u say u ran Web > Server edition, as that only supports 2Gb where XP supports > 4Gb, or was that someone else? It was someone else because STD or ENT are the platforms I am reffering to. > >> They would only benifit if they OS ( kernel ) was using > significantly > >> less machine resources to do they job it did previously > without them. > >> I'd put it to you thats not the case as otherwise MS would be > >> shouting from the houses that 2k3 10% or more quicker at > running all > >> your apps. > > If your talking about applications that are basicly > > single/multi-threaded daemons that require little user > interaction and > > running multipe instances of those daemons, then it's a > different story. > > Why? You should be able to answer that. > >> Again so why isn't MS shouting about this nice performance > increase? > > MS seems to be saying that we should all upgrade to windows XP for > > workstations and 2003 for servers, seems like what one > would do with a > > product that's better than and older version. > > Of course they are they want u to spend money but there's no > big advertising campain touting huge performance increases > that I've seen. > > >> They do? Which? ( I'm talking real work threads here not basically > >> idle threads )? UT for example uses a seperatethread to do DNS > >> lookups but since they are so infrequent event doubling the > >> performance ( which your not doing to see ) would have no > persevable > >> effect on the servers performance. > > Doom3 for one. > > Looks like only one of the 3 threads does any work from the > trace I just did ( like in the case of UT ) could be wrong though. 6 threads there actually then after it finishes initialization, it's 4. > >> Yes you are, no one's saying your not but when you take those > >> opinions and give others advice based on them; when they are > >> unsubstansicated its like chinese whispers. People start > to believe > >> its true just because it was said, even though its not actually so. > > I give you the same advice. > > So what wisper did I start? I thought all my conclusions > where backed up by quantifiable sources. Correct me if I'm wong. Most of mine were as well. So we have different sources obviosly. > > I have agreed with your opinion as long as your not running > more than > > 2-3 game servers, when it gets between 4-16 is where you see the > > benefits. So if you're a GSP(they usually run more than 6 > servers per > > box) or just want every bit of performance you can get(Kid with the > > best stuff on the block) server 2003 will give you what you > want, otherwise XP pro will do fine. > > 16 game servers on one machine either thats pong or I'd hate > to have a server on a machine with that many talk about lag city. > Depends on the machine and configuration, If the machine is correctly configured, it's not Lag City as you put it at all. > Yes we are a GSP one of the biggest in the UK along with > running by far the largest LAN's in the country. We have > tested and continue to test servers across many different > platforms ( one of the very few who have conducted controlled > tests on game server performance across multiple OS's ). I've > seen 0 evidence so far that leads to the conclusion that 2k3 > server is significantly better ( as originally claimed ) than XP. I have seen 0 evidence that 2003 isn't when running a high load. As far as being one of the very few, that maybe the case in the UK. But it is not the case here in the USA. > There is no doubting that there a large amount of improvements in > 2003 over 2000 and also a number of improvements of XP but > I've yet to see or experience anything that justifies > spending �500 on an OS compared with �90. I'd much rather > spend the same money buying another machine which would have > very real performance benefits. > > Steve / K And those are your experiences and what you have seen. I don't doubt what you have seen, but what I have seen & experienced differs very much from what you have. _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
