Multiplayer not in playable state

[ratszo 17]

Well, if you want to test if the game can even handle 100 players with *some* mission, all you have to do is make a mission that automatically deletes all dead bodies and let them go at it in a 50 vs 50 TDM with no extra objects/vehicles/ai. If that doesn't work - Nothing will.

Hi Gal, another aspect of an ideal mission may be optimizing how players 'join-in-progress', JIP.

Missions run much better when a full server is locked..., seems the JIPs suck bandwidth spawning in.

Maybe Bis can look there for improvements.

[NeuroFunker 11]

I can only play with good framrate on servers with a maximum of 30 players on. So it's funny when BIS says this on their official website:

I would love to see if it's even possible to play 120 players on one server.

playing weekly games on russian arma 3 community, about 100-120 players each games, yes you can. It obviously lags at begining until everyone is loaded and whatever else, but then it is very playable. Right now on WOG are 110 people, i think it is playeable if they are still there.

[pils 49]

Ah another issue just came to my mind. Don't know if performance or multiplayer related.

When those 35mm mobile AA vehicles are firing more than just a few rounds, there is some major breakdown of fps. At least for everyone being in the area and I think sometimes even for the one who operates the AA.

So when firing with one of them a few larger bursts towards some town you can be damn sure everyones fps in that town are breaking down to <10 what is really extremely annoying.

[fred41 42]

Will try to provide some profiler for dev. version of DS.

... this are very good news :)

Such a profiler, perhaps based on periodical sampling, could be very helpful for optimizing the script code in a mission.

If coders would be able, to determine exactly, how much time/percent each specific code segment/function need to run,

the performance related quality of missions would be improved a lot.

Even if a simple profiling is already possible, by using diag_ticktime and diag_log and then evaluate the log files, it is not very convenient.

... cant wait ...

Edited November 16, 2013 by Fred41

[terox 316]

Just a heads up on the stress test mission in development

So far it is

52 players (all Arms)

Available params

• Select number of AI (In lots of 50 from 0 to 600)(Unwanted units are not being deleted, they are simply not created via a condition). Deleting them would create a lot of unwanted updating traffic at the start of the mission
  The following options are for FPS comparison *e.g once you know the FPS for a full blown test, restart with these options and see what effect this has on FPS)
  
  • 
    Lock AI waypoints
    Disable simulation on all AI units
    Remove all AI weapons from AI

To create a more consistent test and for better comparison across a number of servers the following steps have been taken

1) Viewdistance is set within the mission (not as a param)

2) There is no randomness with the ai, waypoints etc are all fixed

4) All AI that will be in the mission are present from the start

5) Players respawn, this hopefully will maintain consistent numbers throughout the test

6) No addons, pure vanilla

7) No typical functionality scripts added, eg VAS, patrol scripts, revive scripts etc etc)

There is only 1 looping script, this is serverside and is used for logging data to the rpt once per minute

W.I.P

Vehicle respawn via a killed event (need top be able to consistently transport troops back to the front line, or the number of testing clients will drop)

Positioning of the 600 Ai to give you some fun

There willl be some functions loaded into memory to give a minimal functionality to the mission

For example, creation of Group Icons so that the mission commander knows where the groups are

A loadout script for the vehicles sand maybe the players (not sure if i want to add this)

I'm trying to create as minimalistic a mission as I can but still allow for an enjoyable testing session

The hope is that as BIS address the issues, retesting should see an increase in the number of AI present for the same server FPS.

The mission itself is basically a simple linear clearance of the MSR heading west of the central airbase.

The A.O is split into various phase lines. Each phase line has a set number of AI within it

So for example if you were to find that you could only get a playable fps with 100 AI, the mission would only take you as far as Phase line 1

if 200 AI Phase line 2 etc

This way we can see what effect any fixes that BIS implement have on the playability of the mission.

The end goal being to load all 600 AI with a full house of players.

52 players with 600 AI should be doable without the need for Headless client or AI despawn-respawn scripts

(Am busy over the weekend, but the mission should be avaialble this coming week.

Anyone have any input that I could implement into the mission that would help in the debugging of the issue ?

Edited November 16, 2013 by Terox

[kremator 1065]

^^ Terox if you could do 52 players and 600AI I would be very impressed. Anything over 200AI (without fancy caching) kills fast PCs!

[terox 316]

^^ Terox if you could do 52 players and 600AI I would be very impressed. Anything over 200AI (without fancy caching) kills fast PCs!

Kremator you miss the point of the test mission

The stress test mission is designed to consistently reproduce the low fps that high end servers are getting when running with 30 + players (ideally 40-50 players) without the added clutter of lots of scripts

I do not expect any server to run with 600 AI and have a playable session. If it did, the test mission would fail in its purpose. (This is the reason why the number of AI are selectable as a param

Initially I suspect something in the region of 50 - 150 AI will cause the issue we are all concerned about. (maybe 200 as you say)

Having proved this over a series of servers and collated the information, we can then send all our data to BIS via a ticket.

(The mission will come with a pdf with details how the test should be run and what data should be collected.

For example, BIS may want any of all of the following

Wireshark logs, Freds Arma monitor log, rpt files, bandwidth cfg, server specs etc.

To create consistent testing we need a benchmark mission.

There isn't really anything out there at the moment that we could use, as nearly all large scale missions are cluttered with scripts that could cloud our data and delay BIS's research on the issue

The hope is that as BIS fix or optimise the code, the benchmark mission can then be replayed but with higher AI and achieve a better FPS for AI:Player ratio

Having a benchmark mission with details of server fps's, bandwidth cfg's etc would also be very useful to other server admins who are trying to tweak their servers to their optimum performance

[Von Quest 1163]

following this thead, agreeing with some points, wip ...

Will be patiently waiting to see this improved over time. Slow, but BIS is doing great work.

Getting a tad frustrated myself. Only play MP with one other PC (see sig). We play often

for hours and over time the client pc loses fps to the point of unplayability too much.

Testing and building a Mega-Map with Mods, Scripts, and complete Random Events, and

Waypoints. It's a "Commando-Style" system. Small team vs. the World kinda thing. Whats

weird is the pc's are close, but the host runs fine running the game but the other pc

degrades fps. Then I added the Simulation Manager at 2000 and I noticed a great boost

to frame rates. Same issue on other (client) pc. Continues to degrade over time...

I don't understand why "dead" vehicles eat performance, but not when they are "alive"?

Thinks it would be the other way around. When we play it's very small combat, if any, as

we are black-ops and sneaking around most of the time. Would it be better to set

Waypoints instead of using BIS_fnc_taskPatrol for enemy AI? Shouldn't make a difference

should it? Anyone have good setup/editor tips? ...and what's the best way to setup a simple

Network? Should I use a direct line to connect PCs or keep using a wireless router?

Also just thought; ...is this related to Steam by any chance?

@BIS DEVs

Thanks for following this thread and chiming in. It means a lot to us to know you are active

and care. Too many faceless corporations out there just looking to take your money.

Added:

Forgot to mention. We Desync as well. (Same as FPS?) After some Missions when we check

our watches at the end. He will be 10-20 minutes behind me in time. Ugh.

Edited November 17, 2013 by Goblin
forgot desync note

[aussiebobby 1]

I don't understand why "dead" vehicles eat performance, but not when they are "alive"?

Thinks it would be the other way around.

This maybe nothing or could be something.

I was doing some benchmarks using the helo´s Altis-Benchmark and I noticed and lot of particles emanating from the dead vehicles in the link.

http://img841.imageshack.us/img841/4638/6t6i.jpg

The first time I seem them was when I benchmarked a GT630 and the particles emanating from vehicles like a volcano.I just figured Id overclocked the card a tad to much,and now I seem them with a 7950 OCed.

Ill fit the GT630 and see if I can get a better image of the "volcano like particles"

[Llano 11]

playing weekly games on russian arma 3 community, about 100-120 players each games, yes you can. It obviously lags at begining until everyone is loaded and whatever else, but then it is very playable. Right now on WOG are 110 people, i think it is playeable if they are still there.

On Altis? Could you please go into a town and show me your fps?

[galzohar 31]

In/Out of town or your position in general does not effect your FPS when you are CPU-bound. If you drop FPS when you go in a town you might want to try again in single player with the same settings (keep in mind view distance can be changed by mission scripts and terrain details are completely overridden in MP, defaulting to setTerrainGrid 10) to make sure it's really an MP-only issue.

[zorrobyte 30]

All comes down to the mission.

I hosted 60 slot servers with decent peformance in the past (in alpha)

http://www.youtube.com/watch?v=WBhSXDctY-g

#monitor reports 5FPS but all AI processing is done on the headless client which makes things playable.

[Hud Dorph 22]

When #monitor reports my server 5fps all clients reports unplayable 20fps or less. Maybe its different when using HC.

[gagi 10]

Sure it is !!!

[.kju 3245]

zorrobyte the point is the massive degradation compared to A2/OA.

To make HC use a requirement is hardly the solution either.

[gagagu 21]

@BIS DEVs

Thanks for following this thread and chiming in. It means a lot to us to know you are active

and care. Too many faceless corporations out there just looking to take your money.

Your'e fully right! Thx to devs

[Llano 11]

In/Out of town or your position in general does not effect your FPS when you are CPU-bound. If you drop FPS when you go in a town you might want to try again in single player with the same settings (keep in mind view distance can be changed by mission scripts and terrain details are completely overridden in MP, defaulting to setTerrainGrid 10) to make sure it's really an MP-only issue.

Ehm, if i get you right, you are saying that towns doesn't effect your fps? Because they do

[galzohar 31]

If you were playing with 5 FPS on the server it is not really playable. It will work, sort of, but will be quite laggy if you actually pay attention to anything that actually requires client-server interaction, like seeing actions of other players, shooting them, being in a vehicle they are driving etc.

You can (and often will, though also often not) have perfectly fine FPS and appear to have perfect control over your own soldier/vehicle even when the server has 5 FPS, and in fact even if your internet gets completely disconnected, since those things run on your local machine and only broadcast the results to the server but don't require any reply from the server in order to keep working. That's also why when the server dies you'll notice everyone on teamspeak say "I can play fine but everyone else are running in place/circles/straight lines".

[nicolasroger 11]

Ehm, if i get you right, you are saying that towns doesn't effect your fps? Because they do

You din't get him right. What he mean is that if towns affect your FPS (your case), then you are NOT CPU bound.

[dnk 13]

Client-side I find it fairly unplayable on most servers. It seems to be pure luck when I can and can't get a reasonable game out of a server/mission, and it's not always player count, though certainly that has an impact. Nothing like 2-minute "lag spikes", intermixed with 500-1500ms lag spikes constantly with just 15 people on PvP. Seems most don't have this issue, so it might be due to my high ping (250-300ms generally).

Or I play Domi with a 50ms jitter every other second with just 10 guys on. Doesn't happen in SP. Again, might just be me, and might just be due to my high ping, but I don't recall these issues A2/OA at all. By all means, same ping I'd get pretty good FPS with minimal spikes/jitter on any sized server. Played a lot of Warfare then with 30-40 players and all the AI to go with. It would slow, but never with all this stop-and-go nonsense.

Then there are servers without issue, even with 30 players on sometimes (and other times not). I don't get it.

---------- Post added at 10:34 PM ---------- Previous post was at 10:31 PM ----------

You din't get him right. What he mean is that if towns affect your FPS (your case), then you are NOT CPU bound.

How do you figure that? Seems towns have an awful lot of geometry per scene.

[galzohar 31]

Yes, towns will not affect your FPS if your limiting factor is your CPU. If towns affect your FPS it means your CPU isn't the problem, but rather your GPU is the problem, and thus your problem is completely unrelated to the problem discussed here.

Ping shouldn't have anything to do with your FPS/jitters. It just means you see all network events delayed relative to everyone else, but simulation and graphics should still be calculated at the same speed (they'll just show wrong results due to the delay in your network updates, like players not being in their real position).

[.kju 3245]

I am pretty sure you are wrong there galzohar. Towns means lots of geometry to be calculated by the CPU.

[galzohar 31]

Well, in any case the issue discussed here is low FPS regardless of where you are or what you are looking at. The issue is that in some missions after a while even standing at the end of the map and looking at the sky you will have very low FPS with powerful PCs.

I doubt this has anything to do with location-specific FPS drops in genreal or towns in specific.

In any case whether it's GPU or CPU doesn't matter as long as it's graphics related (that is, if lowering your graphic settings has any effect). After all, the discussed issue simply has nothing to do with graphics and thus anything graphic-related is just a completely different topic/issue.

[terox 316]

The OP is discussing server fps not client fps

The main issue here is server fps, and its inability to run a decent number of AI while hosting a decent number of players in a coop session

. As the server doesn't run graphics, gpu has nothing to do with this and I really don't understand why client side issues are taking presedence in this thread and moving the line of discussion away from the original intention of the OP

Edited November 20, 2013 by Terox

[.kju 3245]

again Terox PvP only is also a main issue, so seems amount of traffic sent by the server