TF2 on 1 vCore VDS - Best configuration and what is reasonab
-
- A semi-regular
- Posts: 17
- https://www.youtube.com/channel/UC40BgXanDqOYoVCYFDSTfHA
- Joined: Tue Nov 09, 2010 11:08 am
TF2 on 1 vCore VDS - Best configuration and what is reasonab
We have recently moved to NFO. We are running TF2 server on a single vCore (which I presume is approximately 1/3 - 1/2 of a real core).
We have noticed that the server FPS can vary wildly from max (depending on the OS, max can be nearly 1000fps) to 20 or lower. We have tried Ubuntu, Gentoo and now Windows 2008 (non-R2).
We have been unable to get reasonably stable FPS at the default 66 tick when the player count goes above 16-ish. Based on the marketing and feedback from sales, we expected the VDS to run 32 players with no problem.
We've resorted to trying various OSes in a plain TF2 configuration (no addons).
So my questions are:
- What's the best OS and related configuration to get a stable FPS? It doesn't even need to be 1000 fps. So far Windows 2008 seems to provide the most stable FPS but it is still dropping at times to < 20.
- CPU seems high at 70%+ with 24 players. Is this normal? As the VDS package was advertised as 1-3 game servers, we were hoping to run two at 24 and 12 or one at 32. However, we can't even get one 24 without FPS tanking.
- Marketing aside, what is reasonable to expect from a 1 vCore running TF2 in terms of max number of players and FPS stability?
As a new customer, I’m concerned that the VDS power/capability is being overstated.
We have noticed that the server FPS can vary wildly from max (depending on the OS, max can be nearly 1000fps) to 20 or lower. We have tried Ubuntu, Gentoo and now Windows 2008 (non-R2).
We have been unable to get reasonably stable FPS at the default 66 tick when the player count goes above 16-ish. Based on the marketing and feedback from sales, we expected the VDS to run 32 players with no problem.
We've resorted to trying various OSes in a plain TF2 configuration (no addons).
So my questions are:
- What's the best OS and related configuration to get a stable FPS? It doesn't even need to be 1000 fps. So far Windows 2008 seems to provide the most stable FPS but it is still dropping at times to < 20.
- CPU seems high at 70%+ with 24 players. Is this normal? As the VDS package was advertised as 1-3 game servers, we were hoping to run two at 24 and 12 or one at 32. However, we can't even get one 24 without FPS tanking.
- Marketing aside, what is reasonable to expect from a 1 vCore running TF2 in terms of max number of players and FPS stability?
As a new customer, I’m concerned that the VDS power/capability is being overstated.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
Yes, I would expect it to. We have many customers who run large servers on their VDSes, and we've done extensive testing to confirm and tweak the performance of these configurations.We have been unable to get reasonably stable FPS at the default 66 tick when the player count goes above 16-ish. Based on the marketing and feedback from sales, we expected the VDS to run 32 players with no problem.
I always recommend Windows 2003 x64. That is what we use internally, and it is solid.So my questions are:
- What's the best OS and related configuration to get a stable FPS? It doesn't even need to be 1000 fps. So far Windows 2008 seems to provide the most stable FPS but it is still dropping at times to < 20.
That does seem a bit high. Is this without plugins? Are you using our default server.cfg?- CPU seems high at 70%+ with 24 players. Is this normal?
We don't really do marketing here in the standard sense. The numbers on our page represent "real world" results based on customer feedback and internal testing, and aren't trumped up. The absolute minimum that you should be able to run on a single core configuration is a single server, and its performance should be comparable to that of a full machine, since the hypervisor does not add significant CPU overhead.- Marketing aside, what is reasonable to expect from a 1 vCore running TF2 in terms of max number of players and FPS stability?
The Orangebox engine is pretty finicky. Have you experimented with "-threads 1" on the command line?
Also, do you know if you're being targeted by any DoS attacks? Especially with TF2 servers, these seem to have become very common recently.
I can take a look at your physical machine as well, if you open up a support request.
-
- A semi-regular
- Posts: 17
- Joined: Tue Nov 09, 2010 11:08 am
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
I will say that NFO is great about responding to questions. That's the main reason we left our last provider.
I've directed the admin for our server that has done most of the testing to this thread. Hopefully he will add to this discussion and/or open a ticket.
I've directed the admin for our server that has done most of the testing to this thread. Hopefully he will add to this discussion and/or open a ticket.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
I'm the admin Zoomer speaks of...
As he said, we have tried Windows 2008 R2, Windows 2008 x64, Ubuntu 10.10, and Gentoo. We get similar results with them all.
The high CPU usage comes with or without plugins/addons... With our plugins/addons loaded, it might make a maximum of 5% difference.
I have not experimented with the -threads option, but I've tried about everything else I can think of and have read about.
This does occur with your pre-allocated server config or our own server config, either one.
I've had no indication on any level that we're under any type of attack.
Here's the basic command line parameters we've been using:
-console -game tf +map pl_badwater +maxplayers 24 -autoupdate -IP iphere -port 27015 +fps_max 0 +exec server.cfg -tickrate 66 +rcon_password somepassword
As he said, we have tried Windows 2008 R2, Windows 2008 x64, Ubuntu 10.10, and Gentoo. We get similar results with them all.
The high CPU usage comes with or without plugins/addons... With our plugins/addons loaded, it might make a maximum of 5% difference.
I have not experimented with the -threads option, but I've tried about everything else I can think of and have read about.
This does occur with your pre-allocated server config or our own server config, either one.
I've had no indication on any level that we're under any type of attack.
Here's the basic command line parameters we've been using:
-console -game tf +map pl_badwater +maxplayers 24 -autoupdate -IP iphere -port 27015 +fps_max 0 +exec server.cfg -tickrate 66 +rcon_password somepassword
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
I don't see any way of editing my post... but, I wanted to say that I just tried with -threads 1 and it didn't seem to make a difference.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
Hmm, it's very odd, then. Having problems with only 24 players in the server is not something that I would expect at all.
Please open up a support request and I'll move you to a different machine, just to be safe about the hardware.
Please open up a support request and I'll move you to a different machine, just to be safe about the hardware.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
When you say that the FPS drops to 20, do you mean that it looks like this:
1000
1000
20
1000
1000
or is it more variable, like this:
1000
500
125
20
500
300
Are you measuring the FPS through net_graph 4 in the client, and is this having a noticeable affect on gameplay?
1000
1000
20
1000
1000
or is it more variable, like this:
1000
500
125
20
500
300
Are you measuring the FPS through net_graph 4 in the client, and is this having a noticeable affect on gameplay?
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
More like:
1000
998
986
1000
20
998
1000
And sometimes:
1000
998
156
20
998
986
1000
We've been measuring it primarily through 'rcon stats'. We've had complaints from some of our regulars, and I have personally had instances like where I shoot a player point blank 3 times in a row and it not register any of the shots.
When we tried setting up a 24 slot TF2 with people playing on it, then tried to set up a private 12 slot TF2 with just 2-3 people on it... we felt massive pauses, rubberbands, and spikes in gameplay; as well as getting numerous complaints from those on the 24 slot TF2 playing.
1000
998
986
1000
20
998
1000
And sometimes:
1000
998
156
20
998
986
1000
We've been measuring it primarily through 'rcon stats'. We've had complaints from some of our regulars, and I have personally had instances like where I shoot a player point blank 3 times in a row and it not register any of the shots.
When we tried setting up a 24 slot TF2 with people playing on it, then tried to set up a private 12 slot TF2 with just 2-3 people on it... we felt massive pauses, rubberbands, and spikes in gameplay; as well as getting numerous complaints from those on the 24 slot TF2 playing.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
Having occasional drops like that means that it's probably being triggered by some specific, infrequent event. Could you try disabling logging, to rule out delays from log writes?
Have you tried running Windows with HPET disabled (under advanced options)? This changes up the timer that it uses. Windows 2003 already doesn't use HPET, so if you're switching to that as a test, you won't need to change this.
Are you running the servers at high priority, and are you logged out of the machine while they run? (Most importantly, you need to not be using VNC, as it is a huge resource hog.)
It might be also be most convenient for you to run net_graph 4 in your client, so that you can correlate whether the dips are causing the problems that you see in-game. If they aren't, it could be that the game is mis-measuring them, and that other factors might be at play.
Have you tried running Windows with HPET disabled (under advanced options)? This changes up the timer that it uses. Windows 2003 already doesn't use HPET, so if you're switching to that as a test, you won't need to change this.
Are you running the servers at high priority, and are you logged out of the machine while they run? (Most importantly, you need to not be using VNC, as it is a huge resource hog.)
It might be also be most convenient for you to run net_graph 4 in your client, so that you can correlate whether the dips are causing the problems that you see in-game. If they aren't, it could be that the game is mis-measuring them, and that other factors might be at play.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
I've tried on both linux and windows, above normal, high, and realtime priorities.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
We've also done the testing without vnc, ssh, remote desktop, etc connected.
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
We'll have to see how the new machine handles first, then, and go from there. If it continues, I'll recommend as next steps (individually, with testing between each, and only continuing if the problem continues):
1. Turning off HPET
2. Turning off logging on the game server
3. Trying Win2k3
4. Trying a managed Linux server (I would have to set this up for you manually), since it uses a specialized kernel and slightly different virtualization setup
1. Turning off HPET
2. Turning off logging on the game server
3. Trying Win2k3
4. Trying a managed Linux server (I would have to set this up for you manually), since it uses a specialized kernel and slightly different virtualization setup
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
FPS meter is not a terribly good measuring method, mostly because its analysis is completely bunk and the graph is misleading; I'd recommend using the tool on your "Managed panel" page if you'd like a graph over time. (Edit: Nevermind that, since you don't have a managed server configuration.)
If their data is accurate, it shows a FPS above 66 except for one reading, which means that you should have seen pretty normal gameplay. Was the server full at the time?
What map are you running? Also, how does this compare to the FPS at your previous host?
If their data is accurate, it shows a FPS above 66 except for one reading, which means that you should have seen pretty normal gameplay. Was the server full at the time?
What map are you running? Also, how does this compare to the FPS at your previous host?
Re: TF2 on 1 vCore VDS - Best configuration and what is reas
The FPS at our previous host, with a massive amount (60+) of plugins loaded, never dropped below like 30 FPS from its maximum.