Servers benchmark

[Karvanoppa 10]

This are old tread, but wanna post here our new server marks.

Name= Olutta.org CO #1

Version 1.59.79548 Linux (Debian 5)

Intel i7 975 Extreme, 12Gb DDR3, Raid5 ("normal" sata2 disks)

Server playing shows smooth and maps changes not show "Wait for host" popup in popular fast game computers.

default arma.cfg

Startup: Extreads=1 Cpucount=4

Results: Min.17fps Avg.18fps Max.19fps

--------------------

Anybody others fighted with virtualized Windows with this? I'm tryed over KVM and have problem with loading times and beachmark are not stable. Normally it is ~20fps but minimun is 12. I think there is something problem with disks speed over KVM, but not relase what i need to change to fix this? I used raw format for KVM images.

Thanks

Edited May 1, 2011 by Karvanoppa

[rexehuk 16]

Surely this test is very limited in scope? I could most likely edit my server config to generate 30-40 fps constant for this test, but the enemy would be absolutely mentally warping because of it. Do you measure anything other than server FPS in order to produce this benchmark?

While server FPS is important, the bandwidth config is second to none in producing smooth gameplay with minimal AI warping, along with good mission design.

Just my two cents. Running your benchmark, our dedi gets around 11 FPS, but I can bet you money that in a normal game ours will have less AI warping than most others who gain higher in these benchmarks. It's not about having 30 groups down in one go, it's about mission design overall, so I don't see this scenario as a good reflection of "decent server configs".

Maybe you need to implement script lag logging and some way to measure the amount of warping going on etc to make it more conclusive.

[nobrainer 0]

NoPryl PLAY1

Homemade

Intel® Core i7 CPU 960 @3.2GHz

12 GB DDR3 PC3-10700

Win7 Pro 64-bit SP-1

A2:CO 1.59 with beta 79600

NoPryl PLAY2

HP Proliant DL380 G4

Dual-core Intel Xeon Processor 2.8 GHz/800MHz-2x2MB L2

6 GB PC2-3200R 400MHz DDR2

Win7 Pro 64-bit SP-1

A2:CO 1.59 with beta 79600

Both servers:

-port=2302 -nosound -world=empty

PLAY1

-cpuCount=4 -exThreads=7 Hyperthreading disabled FPS = 15

PLAY2

-cpuCount=2 -exThreads=3 Hyperthreading disabled FPS = 6

-cpuCount=2 -exThreads=3 Hyperthreading enabled FPS = 6

-cpuCount=4 -exThreads=7 Hyperthreading enabled FPS = 6 all 4 cores used

Note:

This does not apply to PvP servers, only servers running AI. (coop?)

When server FPS drops below 10, then you have no way of telling what the AI does.

Not sure if the AI starts patrolling, moves to waypoints, fire back at you. There seems to be a point between 10-16 FPS where this starts to happen.

Even if a AI "hits" the waypoint, AI moves much less accurate.

Warping is way more noticeable on low FPS servers.

From my point of view, if you are running a mission with AI and the server FPS goes below 16, just restart the mission.....

Edited May 3, 2011 by NoBrainer

[nobrainer 0]

updated last post, but I'm going to test some things on PLAY2 and see if I can conclude on some things

Tested with and without multithreading / hyperthreading....

Edited May 3, 2011 by NoBrainer

[nobrainer 0]

if you disregard warping (network tuning), what can be done to improve FPS on the server?

I can't buy new hardware and I can't change what I all ready have.

I have only A2:CO running on the server.

It's running on 15k SCSI disk in RAID0.

What can be done on the server setup for A2:CO so I get more FPS?

[visceralsyn 10]

Do not run script-heavy missions.

...Syn...

[nobrainer 0]

ok, that means that there is no way of tuning the A2:CO.

Since I'm talking about this missions. This mission don't run any scripts.

So A2:CO is purely based on what the hardware can deliver and how the missions is made.

You can tune A2:CO with respect of network throughput but not on performance.

Well, then I will try to get rid of most of the warping then.

Thanks for reply

[.kju 3244]

Scripts are not doing much if designed well.

AI and player numbers is what counts.

[zorrobyte 30]

27 FPS (17 FPS with two dedicated servers and two clients running, otherwise 23-27FPS on two dedicated servers on same box with remote client concurrently)

I5 2500K @ 5.0GHZ

1333MHZ DDR3

Windows 7 x64

4TB 4x 7200RPM RAID array

http://i51.tinypic.com/2i6c50x.png

Here is a screen 10 minutes in with one client and dedicated server

http://i51.tinypic.com/2uxv229.jpg

Cheers

Edited May 9, 2011 by zorrobyte

[rexehuk 16]

Personally I think the test would be better if it was built as a decent mission with something like

DAC (For AI spawn time measurement - A good indicator if your server is lagged down with scripts) - Good baseline between servers, measure the total time for DAC to finish INIT etc.

Script execution time outputs to RPT on a series of tests running on a loop for a few minutes.

Rather than just 100's of AI down in a single go which doesn't simulate real play at all, just a stupid number of AI that anyone can do with no experience what so ever. Anyone who would make a real mission by just placing FULL AI groups down ALL active even 10KM away should be shot and booted from mission making.

I use DAC as a baseline for my configuration, if I see INIT time go from 40 seconds to 120 seconds I know something is up with the config. I then run some personal benchmark maps to show how much AI warp there is between groups at different distances. Then a few small tests such as picking weapons from an ammo box, is it instant, or is there lag (simple but good indicators).

Edited May 13, 2011 by rexehuk

[LondonLad 13]

[quote=NoBrainer;1910656

<snip>

PLAY1

-cpuCount=4 -exThreads=7 Hyperthreading disabled FPS = 15

PLAY2

-cpuCount=2 -exThreads=3 Hyperthreading disabled FPS = 6

-cpuCount=2 -exThreads=3 Hyperthreading enabled FPS = 6

-cpuCount=4 -exThreads=7 Hyperthreading enabled FPS = 6 all 4 cores used

<snip>

Have you tried it with -exThreads=1 (File Operations)?

I think you only need to run '-exThreads=1' on the server. I read somewhere that the 'File Operations' only need to be multi-threaded (Unless this has now changed).

From previous testing myself I found this the best option across my servers.

[nobrainer 0]

ah!

Cool, will check it out.

[rexehuk 16]

Pretty sure the ExThreads stuff has nothing to do with dedicated servers? It runs on one core only, regardless of what settings you punch in. Only reason I remember running it was due to a bug way back where running without it caused the server to never load.

[nobrainer 0]

Our servers runs on 1 cpu quad core and 2 cpu dual core and it uses all cores.

But that it just what I see.

[rexehuk 16]

We have a quad core and it only EVER runs a total of 1 core (25%) spread over 4 cpus. So while it might run over multiple cores, you don't get the benefit of running 100% x 4, limited to 25% total CPU util.

I hope this changes at some point though with a sexy x64 server EXE for some extra ram usage, and more stuff moved onto the dedi for processing.

[nobrainer 0]

see your point

Thanks

[wangknox 10]

[server Browser Name]

Wardog[TW_Server] by C.T.

[server Specs]

DIY

1 * Intel i7® CPU 920 @ 2.8GHz (4 Physical Processors, 4 Virtual) - Quad Core Processors

18GB RAM 1666 DDR3

1 * 120GB Intel X25-M SSD(G2)

Windows 7 x64 (64-bit OS)

[ARMA Server Version]

a2oa server v1.59.79384

Startup: start /W arma2oaserver.exe -config=server.cfg -port=2302 (no exThreads=x used)

[Test Result ]

During the start (before mission selection): FPS 50 439MB RAM

Once Mission starts it went from FPS 37, to the lowest score of FPS 11 (fluctuated more between 14 & 13)

Ingame (Start): FPS 13, 636MB RAM

RAM started off at 63MB and went up to around 637MB (by the end of the test)

Comment:

PC is not good to be a private server for ARMA

Edited May 17, 2011 by wangknox

[bj2020 10]

OS: windows 7 ultimate 64-bit

OA deticated server: Version 1.59,no addon,no parameter

CPU: Intel® Corei7 CPU 920@2.67GHZ 2.66HZ

CPU Multi-Threading: Disabled (Comment:Until recently (and it's still the case in 1.05), server was only using 2 threads, 1 for main game, 1 for AI ---By Whisper, reference link: http://forums.bistudio.com/showpost.php?p=1661633&postcount=10)

Memory(RAM):6.00 GB

Test 1 (No Overclock)

System BCLK:133MHZ (default)

CPU ratio:Auto 21 (default)

DRAM Frequency:1067 MHZ

fps:14

Test 2 (Overclock 3.486G)

System BCLK: 166MHZ

CPU Ratio:21

DRAM Frequency:1333MHZ

fps: 18

Test 3 (Overclock 3.4G)

System BCLK:200MHZ

CPU Ratio:17

DRAM Frequency:1600MHZ

fps:18

Test 4 (Overclock 4.2G)

System BCLK:200MHZ

CPU Ratio:21

DRAM Frequency:1600MHZ

fps:19-22 (Comment:when cpu die temperature reach threshold, both core clock ratio&voltage will be reduced, that's why fps variable between 19 to 23 frequently.)