Jump to content
Sign in to follow this  
i_deadly_i

SERVER BENCH [Works with HC]

How high was your server?  

7 members have voted

  1. 1. How high was your server?

    • Average 0-10 FPS
      0
    • Average 10-20 FPS
      1
    • Average 20-30 FPS
      1
    • Average 30-40 FPS
      1
    • Average 40+ FPS
      4


Recommended Posts

I proudly present to you the Server Benchmark v019 !!

In this mission some 'greenback' units are defending a hill on Altis from an attack from both the NATO and CSAT forces.

The outcome of the mission seems to vary a lot on the frame rate of the server. From my observations the greenbacks

seem to survive a lot longer when the server frame rate is lower.

This behavior seems to make sense to me as in theory

bullets slow down if the server fps is lower than 30.

'time' variable will be updating slower than real-life time.

Assuming bullets take time to reach a target they should

slow down.

JRFS9ZK.jpg

This is how the benchmark stats will look after it's complete.

The stats are located in the ending screen,

found by clicking on the highlighted arrow.

The Server Bench is a simple mission for server admins to compare performance of different servers and headless client.

You could also use this mission to test out the effects of various AI improvement mods on the server-side frame rate.

Features:

- X Ending stats copied to the clipboard in the end

- X Dumps the ending stats into the server's report file

- X Headless Client Support (* Only supports 1 (one) Headless Client *)

- X Parameters for different several settings

- X AI is spawned "in the briefing screen"

- X Spectator script by Cyrokrypto to check out the firefight

- X Vanilla units on Altis

- X Uses Platinum Patrol by ComfyBlanket for Spawning AI

HOW TO:

1. Run the mission on your server

2. Set the parameters

3. Be sure to enable Headless Client's parameter if running HC!

4. Wait until benchmarking ends or spectate the battle

5. Look at the stats in the ending screen, paste the stats from clipboard or access the stats from the server logs (* uses diag_log, which creates a line in the report file *)

Results :

sMZvSwY.jpg

Here's a comparison of the Broma server vs a dedicated server hosted on a laptop with an i5

If you run the benchmark on your server, be sure to post your stats in the comments!

DOWNLOAD LINKS :

-=v019=-

MEGA

Google Drive

Credits:

Thanks to bohemia interactive for making arma

Thanks to Cyrokrypto for spectator script.

Thanks to ComfyBlanket for ai script.

Thanks to Broma for testing and help.

I don't take any credit for this mission ( I did make it 'tho )

I don't care if anyone re-uploads, edits or breaks this mission...

If you find any bugs, have any suggestions or problems with this mission please be sure to contact me via private message.

Edited by i_deadly_i

Share this post


Link to post
Share on other sites

Thanks to bethesda interactive for making arma

don't you mean Bohemia Interactive? :P

Share this post


Link to post
Share on other sites

Nice to have a counter for the AI so i can check when die avg.FPS begin to be high enough to be useful.

Vanilla System:

95d412a84d291c2537263622a2efcad8.png

With our Repository ( 18GB Heavy, many many Mods and AI Stuff like ASR, VcomDriving, MCC )

bf935754c9b69d7d5fc70921961b07b9.png

System:

Intel i7 4790 // 4x3,6 Ghz (Turbo Boost deactivated )

Host ESXi 5.5 free

Guest WinServer2008R2

I didnt check how the HC support is wörking and whether the AI is completely transferred to HC or not. It was only a quick check! I´m doing later several tests ( 5~10 ). It would be nice to see if you continue to work on this piece, maybe with avg/max traffix send/receive!

Even a multiple HC Support would be nice for cluster-testing!

Pretty, Pretty Useful :) Thank you all for this.

The only thing bothering me is the "remaining AI" factor. Maybe you should force the AI to an level so we can better compare!

Would be nice if some more dudes share their benchs :)

Edited by Astrell

Share this post


Link to post
Share on other sites

Thanks for the feedback :D

 

I didnt check how the HC support is wörking and whether the AI is completely transferred to HC or not. It was only a quick check! I´m doing later several tests ( 5~10 ). It would be nice to see if you continue to work on this piece, maybe with avg/max traffix send/receive!
Even a multiple HC Support would be nice for cluster-testing!

Yes, the paramater will just make the ai script run on the headless client instead of the server.
The problem with multiple HC's would be spreading the ai out, would they switch to the other hc when they reach a certain number or just when the hc's fps drops to less than 40?

Another issue is the group limit, it's at 144. So having more than 1 HC would not benefit that much...
Atleast from the tests I ran I didn't notice 1 HC struggling at all.
Seems like Arma 3's multithreading issue is not that big, although using hc improved my server FPS a lot
in different situations, going from ~10 fps to ~30. (this was on a different mission).
 

 

The only thing bothering me is the "remaining AI" factor. Maybe you should force the AI to an level so we can better compare!
 

This is there to show the diffrence of a slow and fast server. My test showed that running on a slower machine
the amount of remaining ai was alot creater. This means that AI will be more effective and faster when they are running
on a machine that's faster.

Also time slows down when the server FPS gets lower, which means that bullets slow down aswell ( in theory ).

Another great example would be when the group I played with got a dedicated server after having a virtual one
for a while, the AI literally slaughtered us !

 

 

Vanilla System:

95d412a84d291c2537263622a2efcad8.png



With our Repository ( 18GB Heavy, many many Mods and AI Stuff like ASR, VcomDriving, MCC )

bf935754c9b69d7d5fc70921961b07b9.png



 

AI stuff like ASR, VcomDriving will acctually affect the frames more than I suspeceted myself.
Although for smaller scale operations with ~100 AI they are perfect.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now
Sign in to follow this  

×