Recommended server tweaks and how to test
topas
GER Join Date: 2013-10-30 Member: 188872Members
GER Join Date: 2013-10-30 Member: 188872Members
Hi Guys!
i'm new to hosting a ns2 dedicated server and have been working hard on trying to get the best performance out of my hardware. But it is really hard to identify which things i changed gave a better performance and which not.
I really like to test settings BEFORE i enable a server for public audience to give people a good experience.
So how do you test your new settings? I tried to join the server, enabled cheats build everything up to a late game state (and much more like 50 observatorys...) on both teams, put a whole lof of different eggs on the floor, dropped like 100 guns for the marines and added 24 bots to check out the tick rates. Is that the way to do such a test or is there a better possibility like something builtin i missed for a stresstest?
The result of my test: Looks like the bots and the other stuff i dropped needed less performance like real players? I'm not totally sure about this result, can someone verify this?
Let's talk about the perfomance tweaks (i'm in Windows world...)
Set the process priority to high.
I have set the server to use 2 Cores. --> I have a Intel i7 with Hyperthreading, is it better to use one physical core and put the server process to the two HT cores of the physical core, or assign the process to two different physical cores? I have put it to two different physical cores.
Is better to disable the intel autotune that puts the speed down to lower MHz when there is no high load? I disabled it today to see if the server performs better with more stable tickrates.
Is there anything more i can do (don't say OC the i7... it's a non K model)?
I have created a ramdisk drive and put the whole server on it. Don't know if that helps anything...
Cheers
topas
i'm new to hosting a ns2 dedicated server and have been working hard on trying to get the best performance out of my hardware. But it is really hard to identify which things i changed gave a better performance and which not.
I really like to test settings BEFORE i enable a server for public audience to give people a good experience.
So how do you test your new settings? I tried to join the server, enabled cheats build everything up to a late game state (and much more like 50 observatorys...) on both teams, put a whole lof of different eggs on the floor, dropped like 100 guns for the marines and added 24 bots to check out the tick rates. Is that the way to do such a test or is there a better possibility like something builtin i missed for a stresstest?
The result of my test: Looks like the bots and the other stuff i dropped needed less performance like real players? I'm not totally sure about this result, can someone verify this?
Let's talk about the perfomance tweaks (i'm in Windows world...)
Set the process priority to high.
I have set the server to use 2 Cores. --> I have a Intel i7 with Hyperthreading, is it better to use one physical core and put the server process to the two HT cores of the physical core, or assign the process to two different physical cores? I have put it to two different physical cores.
Is better to disable the intel autotune that puts the speed down to lower MHz when there is no high load? I disabled it today to see if the server performs better with more stable tickrates.
Is there anything more i can do (don't say OC the i7... it's a non K model)?
I have created a ramdisk drive and put the whole server on it. Don't know if that helps anything...
Cheers
topas
Comments
You did almost everything you can do.
About the testing result: I can confirm this bots won't raise the network traffic and therefore the load of this modul which causes mostly all rubberbanding etc. in real games.
A thing you could try to test your server install Shine + [Shine] NS2Stats then enable cheats + autobuild + 22 bots (11 vs 11). Watch the load for at least 1 hour. If it stays stable your server is ready for 18 players.
It's more about what my mod does if autobuild is run