UT99 server performance
- UTPe
- Masterful
- Posts: 584
- Joined: Sun Jul 12, 2009 7:10 pm
- Personal rank: Dude
- Location: Trieste, Italy
- Contact:
UT99 server performance
Hi,
I've just finished to setup an ut99 sniper server (SA, camping and also other mods).
I'm not a server expert, I've rebuilt the server from previous configs (mutators + .ini files) so I don't know well where to put my hands to get the max performance. I would like to know something about improving server performance.
Which factors affect it ? the number of the server packages can affect the overall performance ? which are the best net options to get a good server ? a lot of voice packs and custom skins can slow the server (more or less) ? which is the best tick rate ? do others parameters exist somewhere (in UT.ini ?) that can help gamers play better ?
thanks in advance,
Pietro
I've just finished to setup an ut99 sniper server (SA, camping and also other mods).
I'm not a server expert, I've rebuilt the server from previous configs (mutators + .ini files) so I don't know well where to put my hands to get the max performance. I would like to know something about improving server performance.
Which factors affect it ? the number of the server packages can affect the overall performance ? which are the best net options to get a good server ? a lot of voice packs and custom skins can slow the server (more or less) ? which is the best tick rate ? do others parameters exist somewhere (in UT.ini ?) that can help gamers play better ?
thanks in advance,
Pietro
Personal map database: http://www.ut99maps.net
"These are the days that we will return to one day in the future only in memories." (The Midnight)
"These are the days that we will return to one day in the future only in memories." (The Midnight)
Re: UT99 server performance
Performance (CPU or bandwidth) depends on each mod and mutator.
Too many ServerPackages can deter players, unless they are common in other servers, so most of them won't even need to DL.
About tickrate, if your bandwidth is not a problem I'd try something between 30-40, IG and sniper players tend to like servers with higher tickrates.
Just a tweak, at least for the Camp server:
Under [IpDrv.TcpNetDriver]
RelevantTimeout=5.0
Change that to 10 or 15, otherwise, some immobile players at certain positions will become invisible for others (if you played against roof campers on Face you'll know what i mean).
Too many ServerPackages can deter players, unless they are common in other servers, so most of them won't even need to DL.
About tickrate, if your bandwidth is not a problem I'd try something between 30-40, IG and sniper players tend to like servers with higher tickrates.
Just a tweak, at least for the Camp server:
Under [IpDrv.TcpNetDriver]
RelevantTimeout=5.0
Change that to 10 or 15, otherwise, some immobile players at certain positions will become invisible for others (if you played against roof campers on Face you'll know what i mean).
- Feralidragon
- Godlike
- Posts: 5493
- Joined: Wed Feb 27, 2008 6:24 pm
- Personal rank: Work In Progress
- Location: Liandri
Re: UT99 server performance
I am no server admin, so better feedback will come from who actually run popular servers.
However, here's a briefing:
- server packages do NOT affect overall performance. Generally these packages should be retrieved from another server that is not your main UT server (redirect), as the only thing they use (a LOT) is bandwidth;
From there, it's up to the client HDD and CPU speed to uncompress and load them faster or slower, the server is not affected.
- lots of mods can indeed slow down the server, however what you have to check is exactly how much each mod does so. For example, for voicepacks and skins the performance impact is almost null, but if you have an anti-cheat or another CPU/memory intensive mod like a custom gametype or weapon mods, then you should think if you have a good machine for them, what are the best performance options for each one of them, etc.
Generally speaking, is a very bad idea to spam a server with mods, so the less mods the better, so just use the mods you think are absolutely a must.
Also, filling up a server with mods will also greatly affect the time a player has to wait until he/she can actually play, since the player has to download all the mods you added (if the player in question doesn't have them installed nor cached ofc).
- tick rate: you have to find a middle term here. Lesser tickrate is equal to higher performance (since it means the server doesn't have to process things as fast), however a very low tickrate can affect physics simulation server-side (due to the nature of the physics engine), it also means more prediction on the client relative where players actually are in a given moment which can provide a worse experience, etc.
Tickrate of 20 is the default for a dedicated server, but I know plenty of admins use an higher value (way higher in some cases, like 60).
My suggestion is a bit of trial and error, so start at 35 perhaps, but take this value with a bit of salt, as again I don't own any active servers (wait to hear from actual server admins).
- In the ini there are settings on how much bandwidth (max value) each player can have, and the client can also setup this limit (although the max is dictated by the server, the client one is mostly to say to the server that the connection is better or worse, so the server can send more data or less as just the bare essential data).
The exact properties I don't remember exactly (I don't deal with those much), but again, expect a server admin answering that question.
However, here's a briefing:
- server packages do NOT affect overall performance. Generally these packages should be retrieved from another server that is not your main UT server (redirect), as the only thing they use (a LOT) is bandwidth;
From there, it's up to the client HDD and CPU speed to uncompress and load them faster or slower, the server is not affected.
- lots of mods can indeed slow down the server, however what you have to check is exactly how much each mod does so. For example, for voicepacks and skins the performance impact is almost null, but if you have an anti-cheat or another CPU/memory intensive mod like a custom gametype or weapon mods, then you should think if you have a good machine for them, what are the best performance options for each one of them, etc.
Generally speaking, is a very bad idea to spam a server with mods, so the less mods the better, so just use the mods you think are absolutely a must.
Also, filling up a server with mods will also greatly affect the time a player has to wait until he/she can actually play, since the player has to download all the mods you added (if the player in question doesn't have them installed nor cached ofc).
- tick rate: you have to find a middle term here. Lesser tickrate is equal to higher performance (since it means the server doesn't have to process things as fast), however a very low tickrate can affect physics simulation server-side (due to the nature of the physics engine), it also means more prediction on the client relative where players actually are in a given moment which can provide a worse experience, etc.
Tickrate of 20 is the default for a dedicated server, but I know plenty of admins use an higher value (way higher in some cases, like 60).
My suggestion is a bit of trial and error, so start at 35 perhaps, but take this value with a bit of salt, as again I don't own any active servers (wait to hear from actual server admins).
- In the ini there are settings on how much bandwidth (max value) each player can have, and the client can also setup this limit (although the max is dictated by the server, the client one is mostly to say to the server that the connection is better or worse, so the server can send more data or less as just the bare essential data).
The exact properties I don't remember exactly (I don't deal with those much), but again, expect a server admin answering that question.
- UTPe
- Masterful
- Posts: 584
- Joined: Sun Jul 12, 2009 7:10 pm
- Personal rank: Dude
- Location: Trieste, Italy
- Contact:
Re: UT99 server performance
Higor and Feralidragon, thanks for your comments !
any other hint will be greatly appreciated !
any other hint will be greatly appreciated !
Personal map database: http://www.ut99maps.net
"These are the days that we will return to one day in the future only in memories." (The Midnight)
"These are the days that we will return to one day in the future only in memories." (The Midnight)
- Feralidragon
- Godlike
- Posts: 5493
- Joined: Wed Feb 27, 2008 6:24 pm
- Personal rank: Work In Progress
- Location: Liandri
Re: UT99 server performance
In before "why don't I set it to higher values then":Higor wrote: Just a tweak, at least for the Camp server:
Under [IpDrv.TcpNetDriver]
RelevantTimeout=5.0
Change that to 10 or 15, otherwise, some immobile players at certain positions will become invisible for others (if you played against roof campers on Face you'll know what i mean).
- just want to note before someone asks this: the relevancy timeout doesn't count only for players, but for all objects in the game, so the more time you set here, the more bandwidth your server will need, as in theory more objects will be relevant to replicate to the clients in a given moment (this is, if you actually move in the map to actually have more things in view).
-
- Adept
- Posts: 426
- Joined: Tue Feb 21, 2012 7:29 pm
Re: UT99 server performance
Plain and simple UT99 server should be set at a default tickrate (windows 20) & (Linux 25). Why you ask, its simple Epic never opened the pipe on UT99 engine which they could have done if they had actually continued updating but they handed it to UTPG and they made a mess of it. I've been running UT99 servers for about 8 years or more. I can really say less server packages the better and try not to load to many mods they tend to fight for priority. Too many mods can make your server a bit sluggish when you have many clients using them simultaneously. Also make sure the mods you use dont produce errors from green coding and if so make sure to suppress script errors because your ucc.log or server.log will get huge.
I'm sure I have more to add but right now not thinking about it all.
I'm sure I have more to add but right now not thinking about it all.
Re: UT99 server performance
medor wrote:Here the 4 most important set to do.
Netspeed is important. Set a hight and low limiter.
You can see incidences from tickrat 30 with 3 nettspped settings.
http://unrealtournament.99.free.fr/foru ... ?f=9&t=377
Tickrate is important to.
You can see advantage to set it better than 20.
http://unrealtournament.99.free.fr/foru ... p?f=9&t=85
Patch to do.
451.b http://unrealtournament.99.free.fr/utfi ... ch451b.zip
fix tickrate if it' a windows server http://unrealtournament.99.free.fr/utfi ... ch451b.zip
- UTPe
- Masterful
- Posts: 584
- Joined: Sun Jul 12, 2009 7:10 pm
- Personal rank: Dude
- Location: Trieste, Italy
- Contact:
Re: UT99 server performance
thanks again for your comments, very useful !
Personal map database: http://www.ut99maps.net
"These are the days that we will return to one day in the future only in memories." (The Midnight)
"These are the days that we will return to one day in the future only in memories." (The Midnight)
Re: UT99 server performance
My experience is that the more movement the server have to produce the lower the Tickrate, like MH with Feralis Nali-Weapon-mod I can only go as high as 22, but for a slow camping-server or bunnytrack you can go to as high as 30
Re: UT99 server performance
In my experience, it depends on your servers, playerbase and gametype you are running....
A high tickrate won't work well for everyone. Nor will a lower tickrate...the same goes for netspeed.
If you want to run a high tickrate server, your server itself has to be able to handle it. A high tickrate on a low end server will totally obliterate resources and in turn will make all end user experience bad. I do run high tickrate servers, but that's because it works well for me, my playerbase and my servers can handle it. If you are hosting servers that are orientated around instagib, normal weaps etc, I'd suggest a tick of at least 35-45. If you are hosting BT 25-30 is fine. However, even with a "lower" tickrate, you might find some clients are unable to dodge properly for example, and upping the tickrate typically helps that. There's a lot to consider and to type it all out will not cover every possible scenario..
"optimising" a server is not what is best for you. It is what works best for your client base. I'd suggest once your server is up and running, you play around and tweak it slowly. I'm still applying tweaks here and there even since 2006. It's a never ending process, especially if you are hosting dedicated.
A high tickrate won't work well for everyone. Nor will a lower tickrate...the same goes for netspeed.
If you want to run a high tickrate server, your server itself has to be able to handle it. A high tickrate on a low end server will totally obliterate resources and in turn will make all end user experience bad. I do run high tickrate servers, but that's because it works well for me, my playerbase and my servers can handle it. If you are hosting servers that are orientated around instagib, normal weaps etc, I'd suggest a tick of at least 35-45. If you are hosting BT 25-30 is fine. However, even with a "lower" tickrate, you might find some clients are unable to dodge properly for example, and upping the tickrate typically helps that. There's a lot to consider and to type it all out will not cover every possible scenario..
"optimising" a server is not what is best for you. It is what works best for your client base. I'd suggest once your server is up and running, you play around and tweak it slowly. I'm still applying tweaks here and there even since 2006. It's a never ending process, especially if you are hosting dedicated.
-
- Godlike
- Posts: 3774
- Joined: Fri Jan 14, 2011 1:53 pm
- Personal rank: -Retired-
Re: UT99 server performance
This is an old tutorial/explanation written by TNSe but it's still very accurate information:
http://wiki.unrealadmin.org/Netspeed_Tutorial_%28UT%29
One thing I'd like to mention is that everyone is focused on tickrate a little too much. The client's experience depends on a combination of tick, ping, and netspeed. Like baking a cake you can adjust one of the ingredients to make changes but you need to pay attention to the entire recipe to ensure the cake is baked as planned.
http://wiki.unrealadmin.org/Netspeed_Tutorial_%28UT%29
One thing I'd like to mention is that everyone is focused on tickrate a little too much. The client's experience depends on a combination of tick, ping, and netspeed. Like baking a cake you can adjust one of the ingredients to make changes but you need to pay attention to the entire recipe to ensure the cake is baked as planned.
So long, and thanks for all the fish
- Hook
- Inhuman
- Posts: 754
- Joined: Tue Apr 22, 2008 11:21 pm
- Personal rank: UT99 Promoter/Admin
- Location: Minnesota USA
- Contact:
Re: UT99 server performance
The report above by TNSe that JackGriffin pointed out is also what I go by - it seems to be the best info out there, and has been for years.
The cake thing is a good analogy JG!
The cake thing is a good analogy JG!
=Hook=(Member# 626)
HUTP Active Forums: https://hooksutplace.freeforums.net/forum
HUTP UT99 Community Portal: https://hooksutplace.freeforums.net/
OR: https://hermskii.com/hook/ut99_hutp/
UT99 Server -> CROSSBONES Missile Madness {CMM}
* Newest Versions of: PRO-Redeemers | PRO-SNIPER-Redeemers | PRO-SEEKER-Redeemers <-(the Original)
and Now with FOOD FIGHT and Frying Pan arena !!!
IP: 68.232.181.236:7777 <-(NEW IP to come)
UT99 MH Server -> {CMH} CROSSBONES Monster Hunt (MH) by Mars007 (The Original) - IP: 108.61.238.93:7777
HUTP Active Forums: https://hooksutplace.freeforums.net/forum
HUTP UT99 Community Portal: https://hooksutplace.freeforums.net/
OR: https://hermskii.com/hook/ut99_hutp/
UT99 Server -> CROSSBONES Missile Madness {CMM}
* Newest Versions of: PRO-Redeemers | PRO-SNIPER-Redeemers | PRO-SEEKER-Redeemers <-(the Original)
and Now with FOOD FIGHT and Frying Pan arena !!!
IP: 68.232.181.236:7777 <-(NEW IP to come)
UT99 MH Server -> {CMH} CROSSBONES Monster Hunt (MH) by Mars007 (The Original) - IP: 108.61.238.93:7777
- UTPe
- Masterful
- Posts: 584
- Joined: Sun Jul 12, 2009 7:10 pm
- Personal rank: Dude
- Location: Trieste, Italy
- Contact:
Re: UT99 server performance
Thanks all for the comments
We had a lot of troubles about performance so some days ago we decided to remove every mod but SA and camping.
We removed custom skins/voice packs and also other mods (Instagib, UZI + Relics, Telefrag Arena, Reedemer Arena and ChaosUT), including a lot of mutators used within these mods (the ServerPackages list was greatly reduced). This did not help anyway, same problems.
The server is basically a sniper server, classic sniper arena mod + sniper camping with custom rifle/xloc (MapVoteLAv2 installed, thanks Gopo ! ). Anyway a fast SA is much different from slow camping. We observed that a netspeed of 15000/20000 and a tickrate of 30/35 (very high values !) were pretty good for camping but not so much for SA so we have to find a decent compromise to host both of them on the server.
Gopo, I already read the useful doc about tickrate by TNSe, I was able to understand better the problem even though I didn't find a practical solution (parameter=value) for the problem.
One more issue (?): server version is 436 while other current well-known (and well-going) snipers server are all 451. I think we have to upgrade it.
greetings and thanks again,
Pietro
We had a lot of troubles about performance so some days ago we decided to remove every mod but SA and camping.
We removed custom skins/voice packs and also other mods (Instagib, UZI + Relics, Telefrag Arena, Reedemer Arena and ChaosUT), including a lot of mutators used within these mods (the ServerPackages list was greatly reduced). This did not help anyway, same problems.
The server is basically a sniper server, classic sniper arena mod + sniper camping with custom rifle/xloc (MapVoteLAv2 installed, thanks Gopo ! ). Anyway a fast SA is much different from slow camping. We observed that a netspeed of 15000/20000 and a tickrate of 30/35 (very high values !) were pretty good for camping but not so much for SA so we have to find a decent compromise to host both of them on the server.
Gopo, I already read the useful doc about tickrate by TNSe, I was able to understand better the problem even though I didn't find a practical solution (parameter=value) for the problem.
One more issue (?): server version is 436 while other current well-known (and well-going) snipers server are all 451. I think we have to upgrade it.
greetings and thanks again,
Pietro
Personal map database: http://www.ut99maps.net
"These are the days that we will return to one day in the future only in memories." (The Midnight)
"These are the days that we will return to one day in the future only in memories." (The Midnight)
Re: UT99 server performance
medor wrote:For SA you have not ZP ? 103 is cool http://unrealtournament.99.free.fr/utfi ... lus103.zip
- UTPe
- Masterful
- Posts: 584
- Joined: Sun Jul 12, 2009 7:10 pm
- Personal rank: Dude
- Location: Trieste, Italy
- Contact:
Re: UT99 server performance
Yes, we have it
Personal map database: http://www.ut99maps.net
"These are the days that we will return to one day in the future only in memories." (The Midnight)
"These are the days that we will return to one day in the future only in memories." (The Midnight)