Serious Bug: JIP-LAGs (+ possible easy-fix idea)

[Yoma 0]

Of course it will put more strain on server, but again it's just to test the jiplag. I wouldn't quite start running 30 people servers like that...

Has anyone had the chance yet to test it?

[sickboy 13]

With other remote control apps, I have not so much experience, but I think they simply black out the pieces that contain DX stuff.

To run our CoD4 server on our dedi w/2003, DX files HAD to be installed via RD, just so it would load the server console window.

DirectX consists out of a lot more than just Direct3D.

There's DirectDraw, DirectPlay, DirectInput, DirectSound, DirectAlotofThings

So sure, it's possible you needed to install / upgrade directX to run a server.

That doesnt mean that RDP is Direct3D capable

@rundll.exe: Nice find that tightVNC will work. However, can you actually operate the server over normal internet connections, this way? I mean 0.01 fps, doesnt that also mean the server is unresponsive through the remote control session?

[rundll.exe 12]

@rundll.exe: Nice find that tightVNC will work. However, can you actually operate the server over normal internet connections, this way? I mean 0.01 fps, doesnt that also mean the server is unresponsive through the remote control session?

You need a bit of patience... You get an update of the screen every 5 seconds or so... so you can select the mission etc, but it needs some patience... For testing it should work

[dotted 0]

DirectX over RDP works only when the host you are connecting to is a Vista machine (I'm not sure about Win2008, does it have Aero?) - It doesnt work when the target machine is XP or Win2003 doesnt as they use the CPU to render the GUI.

[Yoma 0]

So anyone actually got round to testing this?

[mr.g-c 6]

So anyone actually got round to testing this?

We not yet.... i came out that the guy i mentioned has "only" 512kiloBIT per second.

Also i doubt this would work properly on a slow upload-connection to test if it's really lag-free then.

[-DirTyDeeDs--Ziggy- 0]

I am willing to help,   but I dont know what measurements to take.

I regularly host ArmA with the server exe and join in playing on it.

WARFARE is typically running in persistent mode, so I am guessing I would start the mission with a few players, allow some time for data to grow, then have someone JIP and then check for lag and desynch??

I am a bit confused... Is this hosted with a server exe then joined and played on, or only hosted with client exe?

[Hellop 0]

So anyone actually got round to testing this?

I will test this tonight. Roadrunner cable in the US is 5Mbit down, 1Mbit up.

Hopefully others will also test. The trick is that you need about 12+ people before you see the JIP "red-chain" problem.

In an effort to contribute to the greatly appreciated analysis that Mr.g-c started, I think there may be other issues at play then just JIP lag. From what I have seen using the monitor command, more then 6 players on Warfare 1.1 with a 3000mhz Pentium D (2 processor) with 2GB RAM and 4MB(100MB Burstable) connection, running linux kernel 2.6.18-6-amd64 (64 bit kernel which is recommended for the Pentium D and necessitates installing the ia32-linux package)... then the server FPS drops to under 10 FPS. According to the wiki that is considered overloaded. Any more players and the FPS can go to ~5FPS.

I suspect this is because of the large number of AI present in Warfare, and this is a problem unrelated to the JIP problem. I read in the wiki that 64-bit CPUs are not recommended, but that is not really a viable suggestion in this day and age. Maybe installing a 32-bit kernel would fix this problem. If anyone can post their experiences with server FPS in regards to 64-bit vs. 32-bit kernels, or any other suggestions it would be appreciated.

@Suma, I also think that Actium's 2 points should be considered by the Arma devs:

1. http downloads. i.e. a new command in sever.cfg that is like: httpRedirect=www.myExternalWebServer.com. There you would put your mission .pbos, and ideally, any addon's required to play on that server so that no client would ever get kicked for not having the right files. They would just d/l them and join.

2. Get rid of the whole idea of sending to the clients info about every vehicle that was ever created. The rumor is, that this is done because scripters used the initlines in a createvehicle command to work around limitations of the publicVariable command. This is a hack. It is no longer needed. And you should always start over and do things the "right way" no matter what the cost. Otherwise you just dig yourself a deeper and deeper hole.

I think this thread is a good place for all of us to discuss these suggestions and get input from admins and devs. And of course, if these suggestions are valid, they should also be posted in the Arma2 suggestions forum as Mr.g-c suggested.

[SWAT_BigBear 0]

 If anyone can post their experiences with server FPS in regards to 64-bit vs. 32-bit kernels, or any other suggestions it would be appreciated.

This is where I get dumb founded......

My home pc is all 64 bit.........

Task manager shows several apps running *32.....including an ArmA server, yet it runs the ArmA server much better than our dedi with Server 2003 on a dual cpu AMD 2600 even though the dedi has a 20+ upload,

and my home connection is merely 512kbs up....

So..back to the 64bit concerns......does it or not really matter?

[rundll.exe 12]

-Ziggy- @ June 20 2008,22:01)]I am a bit confused... Is this  hosted with a server exe then joined and played on, or only hosted with client exe?

The test should be performed with your normal arma.exe, and then starting a new MP game from it. (read suma's post again)

[cross 1]

-Ziggy- @ June 20 2008,23:01)]I am willing to help,   but I dont know what measurements to take.

I regularly host ArmA with the server exe and join in playing on it.

WARFARE is typically running in persistent mode, so I am guessing I would start the mission with a few players, allow some time for data to grow, then have someone JIP and then check for lag and desynch??

I am a bit confused... Is this  hosted with a server exe then joined and played on, or only hosted with client exe?

The test is for finding out ;

Potential low/wrong reported FPS to the game engine when the Arma is not in the Focus (not the active screen).

So you need to;

1- Run Arma Game with arma.exe

2- Host a game thru in-game multiplayer/create game/options

3- Have the Arma on the active(focus) screen always (NOT Alt-Tabbed to background)

4- have 10-15 players join in one by one etc..

4- Measure JIP lag/desync repeating Mr.g-c's steps

5- Then send Arma to background by alttabbing and switching to windows and do the measurements again by having people disconnected and re-join the server

These 2 measurements should be different as, due to the potential bug, Arma reports wrong FPS thus uses resources inefficiently  when it is not in the active(focus) screen.

You should expect more lag/desync when Arma is NOT in the focus.

Talk to Mr.g-c on what tools he uses, how he measures etc...

Hope this helps

[Yoma 0]

I know i'm repeating myself but:

Anyone tested it yet? Again i'll be happy to join for testing.

Can't test it myself as i don't have a decent enough upload on my connection.

This is an important issue guys! We need tests!

[suma 8]

What our internal testing seems to show is:

The fix seems to have cured JIP lag for already connected players in all situations except for custom files. When using custom files, there is still a period when the custom files are being broadcasted to all other players, which essentially blocks any other game traffic. This can be observed especially with large counts of connected players. We hope we will bring some solution for this as well, but this is definitely not something you could test with 1.14.

If someone would still like to perform the JIP tests on a non-dedicated host, preferably without any custom files used at all, I guess it would be helpful for us all so that we can be confident the fix really works as expected.

[Yoma 0]

The nice thing is that even if you cannot fix the custom files jip-slowdown, admin can, by disabling/limiting custom files.

I think the amount of positive response when the "mission download fix" is deployed will be simply HUGE.

Again thank you BIS for maintaining Arma.

[mr.g-c 6]

Quote[/b] ]What our internal testing seems to show is:

The fix seems to have cured JIP lag for already connected players in all situations except for custom files. When using custom files, there is still a period when the custom files are being broadcasted to all other players, which essentially blocks any other game traffic. This can be observed especially with large counts of connected players. We hope we will bring some solution for this as well, but this is definitely not something you could test with 1.14.

If someone would still like to perform the JIP tests on a non-dedicated host, preferably without any custom files used at all, I guess it would be helpful for us all so that we can be confident the fix really works as expected.

That is really good to hear Suma!

Quote[/b] ]The nice thing is that even if you cannot fix the custom files jip-slowdown, admin can, by disabling/limiting custom files.

Well yes, but that is not really a solution because according to my experience as player and admin, server-owners mostly & only disabled them so widely because of the LAG they produced. So a solution would be really welcome

Anyway, yes this is probably the most important Fix (for MP) since Arma's start, and i really want to say thank you for finally fixing it (especially if it was rather "simple").

Best Regards, Christian

[aussie dave 0]

Is a new arma_server.exe going to be released, so we can run dedicated servers?

[cross 1]

If they are certain about the fix, i guess they will.

In the meantime, if you own a server and have a decent upload speed, helping out BIS with the testing thru non-dedicated hosting method to confirm bug&fix would be greatly appreciated.

[Hellop 0]

Put up a server. But only a couple people joined. If you see a server called "ARMA JIP-LAG Test", please join.

[mattxr 9]

If no ones actually tested it yet, i will get some clan members to test it. In the next few days.

[cross 1]

If no ones actually tested it yet, i will get some clan members to test it. In the next few days.

matt

we can arrange a testing session GOL & KH together, if you have;

-someone with a decent machine and upload speed.

-necessary tools to measure network traffic per person.

unfortunately i'm missing both but I'll gladly arrange ourside for testing. Maybe followed by an coop event? on a better map than last week's :P.

[ModaFlanker 0]

Find someone with Verizon FiOS [fiber optic] and they should have a kickass upload speed...

[gonk 0]

Peer to peer sharing of custom files like bit torrent might help.  I guess it will not help in this case but AA2..   It might ease the traffic to the server.  as well as the option not to upload ppl's custom files.  Some are just crazy and I would prefer default skins only.

[Orton 0]

I am happy to help i also have these problems i could clean up my server no addons what so ever and start a test.

I always play domination and when people join the lag kicks in and then all threw the game it is a weird lag then it plays real strange.

So when ever you need my help i'll be there or here.

[mattxr 9]

If no ones actually tested it yet, i will get some clan members to test it. In the next few days.

matt

we can arrange a testing session GOL & KH together, if you have;

-someone with a decent machine and upload speed.

-necessary tools to measure network traffic per person.

unfortunately i'm missing both but I'll gladly arrange ourside for testing. Maybe followed by an coop event? on a better map than last week's :P.

Yeah, we had trouble with anyone with decent connection..

Saying that i get 10mb connection in about a month or so.. upgrading it from 4mb free so maybe i can then do it, although BIS might have a release within 2 months who knows

[4 IN 1 0]

how about setting the custom file traffic into a very low priority so that all important things get send first?