SRCDS Steam group


Server Discussion and Dual Cores
#16
afterhoursgaming Wrote:Actually after speaking with a few people who run tf2 servers seems the new patches recently cause lag on servers over 20 slots. Drop your server down 2 slots and see if it lags?

On what platform? My 32 slots are running fine (although i haven't had them 20+ slots filled yet since the update).
Reply
#17
Oen386 Wrote:I did top, and it says 95+%. I assumed that high CPU usage was related to a lack of RAM.

Weird about that server then... I mean if you are claiming a 2.2ghz can run 4 servers with about 70 people. I can not see why mine will not run 1 server with 22 people.

Internet connection is not the issue I am pretty sure, so it has to be the CPU? Thought 2.8ghz would be enough. Sad

What is your connection bandwidth? If this is not a lag introduced by valve as suggested above me then I agree that this sounds like a bandwidth problem.
Reply
#18
Interesting.. so after the recent of patches for the scout?
Reply
#19
Forsaken Wrote:
Oen386 Wrote:I did top, and it says 95+%. I assumed that high CPU usage was related to a lack of RAM.

Weird about that server then... I mean if you are claiming a 2.2ghz can run 4 servers with about 70 people. I can not see why mine will not run 1 server with 22 people.

Internet connection is not the issue I am pretty sure, so it has to be the CPU? Thought 2.8ghz would be enough. Sad

What is your connection bandwidth? If this is not a lag introduced by valve as suggested above me then I agree that this sounds like a bandwidth problem.

My bandwidth is not limited. I am not sure what the max is, but I am also running on WWW2, so people 50-100 miles away can get 5 ping in game.
Reply
#20
Oen386 Wrote:My bandwidth is not limited. I am not sure what the max is, but I am also running on WWW2, so people 50-100 miles away can get 5 ping in game.

And......?

Ping and bandwidth have absolutely ZERO correlation between each other.
Reply
#21
skyride Wrote:Ping and bandwidth have absolutely ZERO correlation between each other.
no. if your data rate comes close to the bandwidth, the ping will rise. this might already happen very early. during my nightly backup to my ISPs ftp server the connection is load at ca. 50% (because I limited the upload) and the ping rises by 80ms or so.

but of course, this does not mean your ping is good if you have a high bandwidth (that's what skyride probably meant).

Oen386 Wrote:I did top, and it says 95+%. I assumed that high CPU usage was related to a lack of RAM.
this might be the cause of the problem. if your ram were limited your system would swap. afaik this would not lead to a high cpu load (maybe high IOWAIT% or perhaps high SYS% but not USER%). either your cpu (you still did not answer what cpu it is, the GHz has no meaning anymore, Xeon is a product line, not a cpu family) is too slow to handle whatever you are doing at your server (bots?), or you have some nasty system settings leading to a high cpu load.

I always say that I would not look at the cpu load, because it does not mean much. But if you are running at 100% and having lags this might nevertheless be a hint to the problem ;-)



btw: yesterday I upgraded my root to 2.6.29-rc4-rt2. now I have seen that srcds has taken more than 100% cpu load some times (up to 200% for a very short time - more is impossible as I have only 2 cores). maybe the assumption that srcds supports only one cpu is incorrect? maybe it only requires a very recent kernel? I am not sure, but unfortunately the interrupt load has rised also with the uprade...

EDIT: That is probably an artefact of top. htop and some other tools do not show this high cpu load. really strange, but it might be due to the fact, that 2.6.29 is still a RC not a release...
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#22
BehaartesEtwas Wrote:but of course, this does not mean your ping is good if you have a high bandwidth (that's what skyride probably meant).

Well, sort of. Your example of an 80ms increase is caused by packet scheduling due to the fact that an ADSL connection is half duplex (i.e. upload and download don't have seperate wires). My point being the bandwidth you have is irrelevant, its how much of it your using that matters.

I just didn't totally finnish off my point, BehaartesEtwas Smile
Reply
#23
skyride Wrote:Your example of an 80ms increase is caused by packet scheduling due to the fact that an ADSL connection is half duplex (i.e. upload and download don't have seperate wires).

Nope, its not DSL, its 100MBit Ethernet full duplex. The ping is a round trip, so both directions matter.

I agree with you, the bandwitdh does not matter if it is large enough. If you are close to the limit, you will get higher pings. If a line is congested, it will take longer to send a packet through it.
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#24
Sorry took me a while to post but, He was running debian 5.0 lenny and was having high CPU usage on servers over 20 slots servers that ran fine before the scout patches were introduced. This may be your issue it may not be. I highly doubt this is a bandwidth issue unless there is a problem with his provider or bad routing. To address the CPU issue he doesn't have a dual core CPU he merely has dual XEON's 2 single core xeons on one motherboard. The xeons are based off the pentium 4 but most likely have higher l2 cache then his pentium 4 machine does they also still used netburst. His pentium 4 box would run the srcds more efficient on a single core application than the lower frequency cpu but there are other factors to include when comparing cpu's.
*Windows lack of output*
You: Hey, I want to run this program!
Windows: Ok.. It crashed... Now what? Give up?
You:...wtf...
*linux output helpful?*
You: ./My_program
Linux:...Failed!...oo kitties!
You:...wtf...
Reply
#25
ah ok. but even on a P4 2.4 GHz 20 slots should run fine (as they already were apparently). maybe he can do a measurment with the fps meter? this would clarify if it is the software or the connection (just to be sure). and maybe I can give some hints how to improve it, if I see the picture. (just post the address of the link which is given below the image)
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply
#26
Well I can't speak on behalf of Oen386 he is the one with this issue. I was comparing his issue to one a friend was having with his server using TF2 his FPS is unsteady but high usually never dips below 900. Behaart your HRG in #sourcekernel? Talk to apoplexy|blaine about the TF2 issues he has been having since the updates perhaps you can find a resolution that will lead to a resolution for Oen386.
*Windows lack of output*
You: Hey, I want to run this program!
Windows: Ok.. It crashed... Now what? Give up?
You:...wtf...
*linux output helpful?*
You: ./My_program
Linux:...Failed!...oo kitties!
You:...wtf...
Reply
#27
afterhoursgaming Wrote:Behaart your HRG in #sourcekernel?
yes, I am. but maybe he should talk to him himself. I dont even have a TF2 server running recently ;-) (it wasnt used at all, so we dropped it)
http://www.fpsmeter.org
http://wiki.fragaholics.de/index.php/EN:Linux_Optimization_Guide (Linux Kernel HOWTO!)
Do not ask technical questions via PM!
Reply


Forum Jump:


Users browsing this thread: 3 Guest(s)