Re: [hlds_linux] OS optimizations for *nix for hlds servers
But the following in your sysctl.conf on Linux and then issues a sysctl -p /etc/sysctl.cnf net.ipv4.tcp_syncookies = 0 # change syncookies to 1 if you will use them net.ipv4.tcp_ecn = 0 net.core.rmem_max = 8388608 net.core.rmem_default = 262143 net.core.wmem_max = 8388608 net.core.wmem_default = 262143 net.ipv4.tcp_rmem = 4096 87380 4194304 net.ipv4.tcp_wmem = 4096 87380 4194304 net.ipv4.tcp_max_syn_backlog = 16384 net.ipv4.tcp_fin_timeout = 15 net.ipv4.tcp_tw_reuse = 1 net.ipv4.tcp_tw_recycle = 1 net.ipv4.ip_local_port_range = 1024 65535 You can also look into using hdparm to turn on ultra dma and writeback cache support on your ide drive , if you use an ide drive. An example is as follows :: hdparm -W1 -c1 -M254 -d1 /dev/hda matt On Monday 22 November 2004 7:17 pm, [EMAIL PROTECTED] wrote: > Anyone have any optimizations/tweaks to the os/network that they want to > share? I thought it might be beneficial to start a thread on it so we could > get input from everyone to see what works / doesnt work. > > -- > Dedicated Servers Available > East Coast Only - 1200gb BW > p4 thru xeon - AIM: gotgameservers > > > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] CPU useage
1000hz is over kill for hlds. As I recall on the game servers we used to run, we found out that 512 hz performed just as well as 1000hz but you didn't see the high cpu usage associated with the overhead of using 1000hz. Just for a side note. 1000hz will produce a substanial overhead compared to 100hz. Matt On Saturday 21 August 2004 3:01 pm, Chris Adams wrote: > Surely 1000Hz on an XP 1700+ is overkill? Especially when you're only using > a ticrate of 200 it seems a bit pointless.. Maybe lower the Hz to 512 or > something? > > Chris Adams > > - Original Message - > From: "Arató Ádám" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Sent: Saturday, August 21, 2004 9:17 PM > Subject: [hlds_linux] CPU useage > > > Hi guys > > > > I have a little problem now. > > My box is an amd athlon xp 1700+, 1024 MB DDR, Asus a7v333, debian linux > > with 2.4.26 kernel. > > Today i recompiled the kernel with 1000 HZ, then i set the sys_ticrate > > value > > > to 200. > > HLDS (cs 1.6) is running, but its CPU usage with 2 players is 99,9%. > > Disabled pingboost and logging, but seems to do nothing. > > > > CPU InOut Uptime Users FPSPlayers > > 99.90 28.98 44.41 6236 100.09 2 > > > > Over 10 players the ping going up to 100-120 ms. > > I'm using the amd optimised binary. > > Server plugins: metamod 1.17.1, adminmod 2.50.59, clanmod 1.82.b1 > > Anybody can help? > > > > a. > > > > > > > > ___ > > To unsubscribe, edit your list preferences, or view the list archives, > > please visit: > > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] After 3 years, I've had enough.
Iam kinda inclined to disagree with your arguement. Valve has made mistakes and are making mistakes now. But some of what your pointing out is too extreme for a company to establish ( come on , using torrents to release patches ) . Right now Valve has alot more on there plate then just us complaining to them. Think about it , Valve was just recently hacked. The code that they have been slaving over, for well over a year was released by hackers. Plus on top of that they are under a tight schedule to reaudit and rewrite most of that code. I'm just hoping HL2 plays out smoother then theese last few months of HL have been. Matt H. On Tuesday 07 October 2003 08:58 pm, 0x2e wrote: > Subject says it all. > > > When people ask serious questions of a companies product reguarding > security, and get no reply, there is something wrong with said company. I > don't buy the idea their network is being rebuilt either. Obviously > something there is still on or this list wouldn't be running nor would > helpvalve@ work. How hard is it to have Leon H. send an email from pine? > > I fully believe M$ was/still has controlling interest in valve. If you > don't understand why that is a bad thing, you never will. > > From the : > > - dick-wagging by Gabe re: DX9 and ATI's "Better support" vs. > OpenGL and NVIDIA, AMD vs. Intel, etc. Look to the archives to > find the truth. > > - their unwillingness to deviate from M$ products (be it mail/web > server used internaly, etc -- Notice I _didn't_ say deviate from > _Windows_ products, nor OS of choice), even if said deviation > would be better. > > - _requiring_ "content" providers to run 2k3 server, but having no > idea why people would be unwilling to "help" them. > > - re-inventing the content distrubution wheel for some > gawdforsaken reason (Sun's Java Webstart vs. M$'s .NET), etc. etc. > Then having no idea why it fell apart. > > - Refusing to listen the outcry from a community they supposedly > "listen to" (re: steam, shields, etc) > > - Not only releasing a buggy as hell BETA product despite serious > problems being found BEFORE the release, But, forcing the > world to download said product without having the infrastructure > to back it. Again, with no clue why it failed. > > Btw valve, where are your buddies at ATI on this? I see they > have no mirror yet that other company you seem to not like has a > few (Nvidia). > > - Refusing to listen again to community suggestions about > possible ways to fix the bandwidth problem (re: .torrent files, > supporting other _established_ mirroring methods like html, rsync, > etc.). M$ way or no way again, Even though it would have > eliminated the problem. > > - Total lack of priorities with again, security. It speaks > volumes when major security patches are produced outside of a > company, well before said company releases their own official > patch. Windows and "It will be included in the next serice pack", > comes to mind. > > > My latest question to Valve was not the only security question that has > went unanswered. A few weeks back I pointed out how the SteamID, smelled > like a passport ID for example. I have never seen a reply from valve on > that nor many other concerns. Btw, if you don't know why passport is BAD, > google it. It has little to do with the fact that M$ runs it. > > > I could go on but I won't. You guys get the picture heh. > > I have removed my CS servers, and replaced them with UT2k3 and Q3A. You > wanna see SWEET games? Play 'em :-) .. Native linux binaries too! > > For UT2k3 at least, it's VERY easy to code your own mods. By mods I don't > mean like CS/DoD-type a deal where it totaly changes the game (which is > still very easy). I mean admin stuff, changing weapon skins/power, etc. > Basicaly Metamod, but supported natively! (through UTScript) > > Unlike HL where it's a mod to the server that even though being used on > 99.9% of every server out there, Valve refuses to add it as an official > feature (doesn't come with HLDS). Amazing eh? > > > Happy trails peeps, been fun. > > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] Vivendi Universal confirms delay due to code theft
Ive gotten word that the beta of this game is already out from a very reliable source.. Matt On Tuesday 07 October 2003 09:15 am, Tony Di Schino wrote: > http://money.cnn.com/2003/10/07/technology/vivendi_code.reut/index.htm > > "It's serious because it forces us to delay the launch of the game by > at least four months, that is to April 2004. Just the time to rewrite > parts of the game." > > > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] The CPU usage issue.. what should I go with: Linux or Win32?
But it runs at 5 % omg.. its true I tell ya .. true =P On Sunday 28 September 2003 08:48 pm, James Sykes wrote: > Lets not deny that this release is using A LOT more cpu than the last. > Its pretty much FACT - unless you happen to be one of the lucky few with > a SUPER KERNEL that runs 5% cpu with a 32 player serverand we all > know that's bs. > > -Original Message- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Britt > Priddy (PZGN) > Sent: 29 September 2003 03:33 > To: [EMAIL PROTECTED] > Subject: Re: [hlds_linux] The CPU usage issue.. what should I go with: > Linux or Win32? > > exactly what m0gely said. > > less CPU for me on AMD 2800+ > > > - Original Message - > From: "m0gely" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Sent: Sunday, September 28, 2003 9:11 PM > Subject: Re: [hlds_linux] The CPU usage issue.. what should I go with: > Linux > or Win32? > > > James Sykes wrote: > > >>If the new linux HLDS is really doubling CPU usage over 3.1.1.0c > > then > > > > how > in the world can I host enough games to cover the cost of my > > > server? > > > > > > You don't :) > > > > At 14 players in de_aztec, I don't see anywhere near double the CPU. > > It's > a > > > little more, Double though? C'mon. Right now I have 14 players in > > de_inferno > > > and it's 38% on my dual AMP MP 2400+. This is with MM 1.16, VAC, HLG > > 1.5, > AMX > > > 0.9.3 (stats disabled) and Statsme 2.7.1. This is hardly any > > different > then > > > before w/ 3.1.1.0c. It's not good, but not really any different. > > > > -- > > - m0gely > > http://quake2.telestream.com/ > > Q2 | Q3A | Counter-strike > > > > > > ___ > > To unsubscribe, edit your list preferences, or view the list archives, > > please visit: > > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: > http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > > > > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
Re: [hlds_linux] HOW TO GET SUPER LOW CPU USE!! THANK DLINKOZ
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 2.2 uses a different vm, and shouldn't be used, its old its obselete and i would be surprised if any new hardware worked -good- on it. .. As for 2.4.9 the vm that was in there was removed in all kernels versions after it.. I'm not sure the reasoning why , but i believe Linus was having difficulities with it .. ie he had poor io and interactivity performance from it. So he ripped it out and put in a vm based off of the earlier vm work in the aa patchset for the 2.4 kernel tree's. As for the vm in 2.4.9 , it's no longer worked on. The author of that particular vm Rik Van Riel, has a newer vm out based off of reverse mappings ( not sure how this works , but you can check out his site http://www.surriel.com/patches/ ). Anyhow the rmap vm he made is now used in all newer redhat kernels, and a light implementation of it was merged into 2.6. It would be more intresting to benchmark the 2.4.9 kernel, against a more modern kernel like 2.4.22 ( applied w/ either aa-vm or rrmap-vm ) and then with 2.6.0-testX version. However I still believe most of this is Valve's fault.. They either need to shape up and improve the cpu usage of there products, or all together release the source code for other ppl to do work on it. This is ridiculous that server administraters have to downgrade there systems in order to get better half-life server results. Matt On Monday 15 September 2003 09:56 am, Sindre wrote: > could anyone confirm if 2.2 kernels work as good as 2.4.9? > > - Sindre > > >= Original Message From "Kevin J. Anderson" > > <[EMAIL PROTECTED]> > > = > > >->-Original Message- > >->From: [EMAIL PROTECTED] > >->[mailto:[EMAIL PROTECTED] Behalf Of Daniel > >->Stroven > >->Sent: Monday, September 15, 2003 1:37 AM > >->To: [EMAIL PROTECTED] > >->Subject: Re: [hlds_linux] HOW TO GET SUPER LOW CPU USE!! THANK DLINKOZ > >-> > >-> > >->Those #'s look awesome, but for security purposes, 2.4.9 is not really a > >->kernel I want running. As pointed out by my friend matt, the > >->difference in > >->2.4.9 from 2.4.10 and higher is the VM used. But exploits like ptrace > > and ->others could make it vulnerable to remote exploits. We are going > > to test ->the kernel on the box to see results of usage. But, I doubt we > > ->will keep it > >->if we can not make it extremely secure. > >-> > >->dan > > > >You could probably get away with a debian install w/ the 2.2.x kernels > > that are still being kept secure. > > > >kev > > > > > >___ > >To unsubscribe, edit your list preferences, or view the list archives, > > please > > visit: > >http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > ___ > To unsubscribe, edit your list preferences, or view the list archives, > please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux -BEGIN PGP SIGNATURE- Version: GnuPG v1.2.3 (GNU/Linux) iD8DBQE/ZidbleY/n9G/oZ8RApGCAJ48did2MQvfsag4JCObMF3uGxwY8gCfaoy4 zSjDD8OvM5GBZcpSXP/F9pA= =NK5O -END PGP SIGNATURE- ___ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux