Dedicated Server - FPS Tests & Results

SERVER SPECS

Processor:

Quad Core Intel Xeon X3220, 4 x 2.50GHz

RAM:

8GB ECC 667MHz Ram

Network:

Dedicated full duplex 100Mbit

Average / Max Player Count:

15 / 36

Min / Max server FPS:

5 / 35

Mission Used:

Altis Insurgency

Mods Used:

@CBA_A3, @JayArma3Lib, @ACRE, @AMS (to monitor server stats)

CONFIG

MaxMsgSend=1024;
MaxSizeGuaranteed=1024;
MaxSizeNonguaranteed=64;
MinErrorToSend=0.0040000002;
maxCustomFileSize=0;

MinBandwidth=7864320;		// 512kbps 	x 	avg. 15 players (into bps)
MaxBandwidth=18874368;		// 512kbps 	x 		 36 players (into bps)

STATISTICS

Initial server start with 4 players

Server FPS: 32

CPU: 10%

I/O: 84kbps / 449 kbps

Hitting a town with 8 players

Server FPS: 15

CPU: 22%

I/O: 128kbps / 841 kbps

21 players online

Server FPS: < 5

CPU: 27%

I/O: 145kbps / 2056 kbps

Server idling after running for a few hours (no players connected)

Server FPS: 11

CPU: 28% (why?)

Memory usage peaks to about 1.77GB.

At server start the average server FPS is 35. As soon as we hit a town (and the AI spawn), the server FPS drops to 15 (with 8 players connected). With 21 players online, the FPS drops to < 5 FPS.

What confuses me the most is why the server is running at 11 FPS and using 28% of the CPU when just idling. There should be no AI (since AI despawn when no players are present), so what is it calculating?

I've read this guide but the changes made appear to make little difference.

It seems clear that the server isn't being limited by hardware since its barely using the resources it has available, which leads me to believe either the network is bottle-necking or Arma 3 isn't taking advantage of the server's resources.

Is this the expected frame rate/results for a server with these specs? What framerates do other server administrators experience?

the subject of server optimisation has never been clearly explained to us. There are a lot of questions that remain unanswered and have been for over 10 years now.

That tutorial link, is not actually a tutorial, but an attempt to raise those questions and get a better understanding so that it can eventually become a tutorial. It has so far failed to achieve that.

the subject of server optimisation has never been clearly explained to us. There are a lot of questions that remain unanswered and have been for over 10 years now.

That tutorial link, is not actually a tutorial, but an attempt to raise those questions and get a better understanding so that it can eventually become a tutorial. It has so far failed to achieve that.

Its ridiculous to think that something so crucial to server optimisation has not already clearly explained, surely this is the type of information server admins first look for when setting up their server (and according to you, the community has been in the dark for over 10 years! :eek:).

Although your post wasn't a tutorial, I still found a large amount of the information useful. I appreciate your efforts for attempting to get solid answers on the matter. I fail to believe that nobody has any information regarding this, the developers must have answers to these questions so why are they ignoring it?

Is this the expected frame rate/results for a server with these specs? What framerates do other server administrators experience?

... the length of a simulation cycle (fps is the reciprocal of that) depends of a lot factors.

Server hardware performance, mainly CPU clock and architecture, amount of physical memory and io performance, is essential,

but other details like, server OS, swapfile configuration, network configuration params are important too.

Very often the meaning of mission "quality" is overlooked. Sometimes, ONE little line of bad code or the absense of some important lines (cleanup routine),

can ruin your performance (fps) totally.

I see you use @ASM, so you are not fully in the darkness. With a bit experience, ASM can help you, to spot problems caused by your mission code.

The performance impact from number of current alive AI, can be very different, dependent of the state of AI (just patrolling, alerted, fighting).

You see it is not simple to answer your question, but i hope ASM will help you a bit.

Greets,

Fred41

SERVER SPECS

Processor:

Quad Core Intel Xeon X3220, 4 x 2.50GHz

RAM:

8GB ECC 667MHz Ram

Network:

Dedicated full duplex 100Mbit

Average / Max Player Count:

15 / 36

Min / Max server FPS:

5 / 35

Mission Used:

Altis Insurgency

Mods Used:

@CBA_A3, @JayArma3Lib, @ACRE, @AMS (to monitor server stats)

CONFIG

MaxMsgSend=1024;
MaxSizeGuaranteed=1024;
MaxSizeNonguaranteed=64;
MinErrorToSend=0.0040000002;
maxCustomFileSize=0;

MinBandwidth=7864320;		// 512kbps 	x 	avg. 15 players (into bps)
MaxBandwidth=18874368;		// 512kbps 	x 		 36 players (into bps)

STATISTICS

Initial server start with 4 players

Server FPS: 32

CPU: 10%

I/O: 84kbps / 449 kbps

Hitting a town with 8 players

Server FPS: 15

CPU: 22%

I/O: 128kbps / 841 kbps

21 players online

Server FPS: < 5

CPU: 27%

I/O: 145kbps / 2056 kbps

Server idling after running for a few hours (no players connected)

Server FPS: 11

CPU: 28% (why?)

Memory usage peaks to about 1.77GB.

At server start the average server FPS is 35. As soon as we hit a town (and the AI spawn), the server FPS drops to 15 (with 8 players connected). With 21 players online, the FPS drops to < 5 FPS.

What confuses me the most is why the server is running at 11 FPS and using 28% of the CPU when just idling. There should be no AI (since AI despawn when no players are present), so what is it calculating?

I've read this guide but the changes made appear to make little difference.

It seems clear that the server isn't being limited by hardware since its barely using the resources it has available, which leads me to believe either the network is bottle-necking or Arma 3 isn't taking advantage of the server's resources.

Is this the expected frame rate/results for a server with these specs? What framerates do other server administrators experience?

try running altis but completely empty , have some friends joining you and enter one of the bigger towns and measure your performance then.

p.s the performance of A3 is based on core 0 so the higher the clock/architecture the better the performance

which means those CPU% are misleading as that is an AVG across the cores meaning core 0 is probably close to running at %100 while the other 3 are doing ver y little hence the performance wall you are hitting.

have a look at core 0 when performance starts to tank.