Window Vs Linux
Freak
Join Date: 2002-10-31 Member: 1811Members
Join Date: 2002-10-31 Member: 1811Members
I was wondering if anyone has done any bonefied performance testing between the two O/S as far as server performance. I switched my 26 player DOD server to Linux thinking I would have a performance gain and is was a horrible change. My avg ping running Linux was 120-150 and when I went back to W2K, it dropped all the way to 90. The DOD server is a P4 1.7g w'512 DDR. I am going to switch my 20 slot NS server to W2K and see what happens. The NS server is a P4 2.53G w/512meg DDR. My avg ping is now 150-180. I will post the results.
Any other experienced server ops have input on this subject?
Any other experienced server ops have input on this subject?
Comments
Linux isn't as hoggy as Windows, so people can run it on slower boxes. We ran a 10-player CS server on a 550MHz <i>no problem.</i>
The server calculates where each bullet goes when you fire...same with turrets. NS has a lot more entities than CS or TFC. No turrets in CS, one turret per player (as an Engineer) in TFC, <i>unlimited</i> turrets in NS...do the math.
None of this is new information nor does it explain the apparent view that the Linux server runs slower than the Windows server <b>on the exact same machine</b>.
None of this is new information nor does it explain the apparent view that the Linux server runs slower than the Windows server <b>on the exact same machine</b>.<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
I was <b>clearly</b> NOT replying to the original post. <b>I was answering another question.</b> Note the quotebox in my post.
The server calculates where each bullet goes when you fire...same with turrets. NS has a lot more entities than CS or TFC. No turrets in CS, one turret per player (as an Engineer) in TFC, unlimited turrets in NS...do the math. <!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
maybe i'm confused. but i don't see how any of this is relevent. considering freak can run linux or doze on the same box and get the better perf in doze is jacked up IMO.
The original post is "which is better" : doze
My question was why?: no good answers yet except <!--QuoteBegin--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> </td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->the code not being optimized yet.<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->. Which is only half true. because with every compliler comes flags....and those flags (believe it or not) are optimization flags. hhmm i wonder what those do?? <!--emo&:0--><img src='http://www.unknownworlds.com/forums/html/emoticons/wow.gif' border='0' valign='absmiddle' alt='wow.gif'><!--endemo-->
Oy jaysus forking tap-dancing ke-rist. Let's look, shall we?
The portion you quoted:
<!--QuoteBegin--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> </td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->
i'm wondering what the deal is. lately linux has be kicking butt and taking names, but now these new mods are coming out and trashing hlds_l performance. Is it because the programmers (of the mods) do not know how to optimize code for linux? they not read thier gcc books? i seriously think linux is getting the short end on HLds. but its not valves fault, just the mod programmers i believe. correct me if i'm wrong.<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
Note that he is making a comparison. "Trashing hlds_l", "optimize code for linux" and "linux is getting the short end". He is clearly making the same comparison as the original poster which is "Linux appears to be slower than Windows".
Your response didn't answer that. No, it answered the age old question "Why is NS slower than [insert mod here]". You'll note that he didn't ask that question. In fact one of the comparison statements includes other mods, "these new mods are coming out and trashing hlds_l performance." We know why NS takes more resources <b>in general</b> than other mods. That wasn't the question by the original poster nor this poster. The question was why it seems to be slower on Linux than on Windows.
Brainiacs like you are why we have debates over the first ammendment in the US. You like to read a portion of a statement and ignore the rest, respond and think you're addressing the whole statement. Read it <b>all</b>, comprehend, then reply!
LOL, love it!!
14 player NS server 212.149.105.64:28000
10 player NWN server 212.149.105.64:5121 (not sure about port number)
16 player JK2 Server 212.149.105.64:28070
10 player JK2 Server 212.149.105.64:27070
10 player JK2 Server 212.149.105.64:29070
Those servers are running on same machine and same time without any lag or ping problems.
AMD 2000 XP+ with 512 ddr and 100Mbits connection.
Redhat 7.2 Linux
ps xuf command shows this:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
hl 6948 0.0 0.2 2500 1324 pts/1 S 07:58 0:00 -bash
hl 7028 0.0 0.1 2636 672 pts/1 R 08:35 0:00 ps xuf
hl 6014 0.0 0.2 2640 1180 ? S Dec12 0:00 SCREEN -mdS jk2 ./jk2ded +set dedicated 1 +set fs_game admi
hl 6015 1.2 7.3 95160 37820 pts/2 S Dec12 10:17 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 4114 0.0 0.1 2640 904 ? S Dec11 0:00 SCREEN -mdS nwn ./nwserver -module Chapter1 -dmpassword nwn
hl 4115 10.4 5.3 130868 27800 pts/5 S Dec11 297:58 ./nwserver -module Chapter1 -dmpassword xxxx -playerp
hl 4116 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx -pla
hl 4117 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4118 0.0 5.3 130868 27800 pts/5 S Dec11 0:04 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4121 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4109 0.0 0.2 2640 1116 ? S Dec11 0:00 SCREEN -mdS jk3 ./jk2ded +set dedicated 1 +set fs_game admi
hl 4110 1.0 7.9 90368 40836 pts/4 S Dec11 28:26 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 4105 0.0 0.2 2640 1116 ? S Dec11 0:00 SCREEN -mdS jk22 ./jk2ded +set dedicated 1 +set fs_game adm
hl 4106 0.1 6.4 90344 33480 pts/3 S Dec11 4:18 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 1074 0.0 0.4 3952 2280 ? S Dec09 0:27 /usr/bin/tcl /home/hl/halfd/halfd -g ns -e 28000
hl 6111 46.4 20.6 112284 106276 ? S Dec12 340:22 ./hlds -game ns -port 28000 -pingboost 3 +sys_ticrate 10000
14 player NS server 212.149.105.64:28000
10 player NWN server 212.149.105.64:5121 (not sure about port number)
16 player JK2 Server 212.149.105.64:28070
10 player JK2 Server 212.149.105.64:27070
10 player JK2 Server 212.149.105.64:29070
Those servers are running on same machine and same time without any lag or ping problems.
AMD 2000 XP+ with 512 ddr and 100Mbits connection.
Redhat 7.2 Linux
ps xuf command shows this:
USER PID %CPU %MEM VSZ RSS TTY STAT START TIME COMMAND
hl 6948 0.0 0.2 2500 1324 pts/1 S 07:58 0:00 -bash
hl 7028 0.0 0.1 2636 672 pts/1 R 08:35 0:00 ps xuf
hl 6014 0.0 0.2 2640 1180 ? S Dec12 0:00 SCREEN -mdS jk2 ./jk2ded +set dedicated 1 +set fs_game admi
hl 6015 1.2 7.3 95160 37820 pts/2 S Dec12 10:17 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 4114 0.0 0.1 2640 904 ? S Dec11 0:00 SCREEN -mdS nwn ./nwserver -module Chapter1 -dmpassword nwn
hl 4115 10.4 5.3 130868 27800 pts/5 S Dec11 297:58 ./nwserver -module Chapter1 -dmpassword xxxx -playerp
hl 4116 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx -pla
hl 4117 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4118 0.0 5.3 130868 27800 pts/5 S Dec11 0:04 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4121 0.0 5.3 130868 27800 pts/5 S Dec11 0:00 \_ ./nwserver -module Chapter1 -dmpassword xxxx
hl 4109 0.0 0.2 2640 1116 ? S Dec11 0:00 SCREEN -mdS jk3 ./jk2ded +set dedicated 1 +set fs_game admi
hl 4110 1.0 7.9 90368 40836 pts/4 S Dec11 28:26 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 4105 0.0 0.2 2640 1116 ? S Dec11 0:00 SCREEN -mdS jk22 ./jk2ded +set dedicated 1 +set fs_game adm
hl 4106 0.1 6.4 90344 33480 pts/3 S Dec11 4:18 ./jk2ded +set dedicated 1 +set fs_game adminmod +exec serve
hl 1074 0.0 0.4 3952 2280 ? S Dec09 0:27 /usr/bin/tcl /home/hl/halfd/halfd -g ns -e 28000
hl 6111 46.4 20.6 112284 106276 ? S Dec12 340:22 ./hlds -game ns -port 28000 -pingboost 3 +sys_ticrate 10000<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
Doesn't say anything. Just shows your running multi servers with no one on any of them. as the only one that isn't idel and has cpu useage is NS. And thats 46% of the cpu. still meaningless, without knowing how many players are one it. Lets see that when ever one of thoose servers are full. Because they will all be lagging like mofos.
Usually my servers are most popular and they are full about 12 hours per day. And pings are 15-30 for finnish players.
NS server cpu max usage is 30-35% and when server 18 player server cpu max usage is 35-45 %.
12.12 time 20:26
hostname: Hukkalandia NS v1.03 #1 (#hukkalandia @qnet)
version : 46/3.1.1.0 2163 insecure
tcp/ip : 127.0.0.1:28000
map : ns_eclipse at: 0 x, 0 y, 0 z
players : 14 active (14 max)
# name userid uniqueid frag time ping loss adr
# 1 "HuntaX" 1 xxxx 0 03:25 94 1 213.250.118.250:21040
# 2 "nahaaka" 2 xxxx 1 03:23 20 1 62.248.246.20:27005
# 3 "$ Korpisoturi Ahola" 3 xxxx 0 03:17 30 0 80.222.219.81:27005
# 4 "KekeLoordi" 4 xxxx 2 03:08 28 0 80.222.12.245:21029
# 5 "SuPeR" 5 xxxx 0 02:57 71 0 212.88.81.5:28378
# 6 "Erwin" 6 xxxx 1 02:52 76 0 212.50.208.211:27005
# 7 "n00b???YES" 7 xxxxx 0 02:37 31 0 80.74.194.140:52229
# 8 "Macce" 8 xxxx 0 02:11 38 0 80.222.208.93:27005
# 9 "digg" 9 xxxxx 0 01:53 48 0 213.141.100.100:27005
#10 "HawkinS" 10 xxxx 0 01:18 92 3 80.74.208.221:27005
#11 "kyntteri" 11 xxxx 1 01:06 59 0 80.223.173.210:27005
#12 "Merihobu" 12 xxxx 0 00:51 37 0 80.222.73.111:27005
#13 "fibbe" 13 xxxx 0 00:39 36 0 80.222.70.204:27005
#14 "Kinkki" 14 xxxx 0 00:21 186 70 80.222.67.123:27005
14 users
NS server cpu max usage is 30-35% and when server 18 player server cpu max usage is 35-45 %.
<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
plz tell me your secrets! because i have an AMD2000+ with 640mb pc133 (ddr doesn't make THAT much of a difference!) ....i can only get your claim on ns_eclipse....hows your hera?
back to freaks and my original questions:
i'm wondering what the deal is. lately linux has be kicking butt and taking names, but now these new mods are coming out and trashing hlds_l performance. Is it because the programmers (of the mods) do not know how to optimize code for linux? they not read thier gcc books? i seriously think linux is getting the short end on HLds. but its not valves fault, just the mod programmers i believe. correct me if i'm wrong.
There are several answers for several different questions, here. I simply filled in for ONE of them. Is that so terrible?
Okay, okay. It's just that the tone of his reply was insulting, and I've done nothing to initiate it, so we got into a little argument...that's it. <!--emo&:p--><img src='http://www.unknownworlds.com/forums/html/emoticons/tounge.gif' border='0' valign='absmiddle' alt='tounge.gif'><!--endemo-->
<b>greydmiyu</b>, you didn't have to be so rude, and I probably shouldn't have replied.
ehh, DDR <b>does</b> make a difference on a newer athlon. You wouldn't have noticed it on a t-bird, and maybe not on an earlier palomino (like xp 1500-1800) but the core clock is fast enough in the 2000 that it's now necessary. Your pc133 is creating a bottleneck that slows down your performance. I'm not saying you need DDR400; 266 (pc2100) should do just fine, and it's getting cheaper all the time.
Also, if you're running linux, make sure you have as much swap space as you possibly can, preferably on a different physical disk than the one running HLDS. I've found that to improve performance by a huge amount.
Back on topic: Flayra didn't have time to optimize the linux code as much as he would have liked to. Those of you who were around pre-release and read the playtest changelogs should know this. I think he plans to finish the optimizations in the next patch, provided he doesn't have to deal with too many more balance issues.