multiplayer limits with 3dgs

Pardon me Doug but I dont feel that answer was a very fair response to the question. For years people have been trying to accomplish varying degrees of MMO type 3D interaction with 3DGS and no one has been succesful with more than 2-4 players online.. That is using the most basic data objects with a simple shape for the player and no terrain or models to clutter up the rendering pipeline..

3DGS uses DX net code for their network layer which has been proven unstable in the past and is no longer supported by Microsoft. Every player who has attempted any kind of multiplayer connectivity with 3DGS beyond 4 players has been forced to go to a third party network layer. This fact, and the fact that trying to get support on the native network code was so difficult in these forums are the reasons my project had to switch to a more flexible engine. This after we had invested over $3000 in commercial team licensing and personal Pro editions as needed.

@Prence: You can continue making your game with 3DGS, only you can decide if it meets your criteria or not, but try some searches in the forums here for populace and Raknet.. I know nothing about this Populace plugin but supposedly it has realistic and useable 3d many player support.. Raknet is a very viable third party network app written by a very competant network GURU who has integrated into many commercial systems. The down side to Raknet is you will have to get into the C++ stuff in order to tailor it to your needs.

Truth be told Fastlane claims to have been working on getting a many player "demo" of what 3DGS can do for over 3 years now. I believe in one of his more recent posts he claims his university is funding his "analysis" of the 3DGS connectivity and he has 2 full time interns working on it.. To my knowledge, after 3 years of solid work, including professional funding and interns all he has been able to do is run a 3DGS app in "server" mode that allows approx a couple hundreds "connected" clients but without the necessary 3D objects required for an MMO. (EI: In plain english the client software "connects" to the server software but does nothing with the connection as in there is no 3D object handling, movement or any kind of interaction other than the initial connect between the clients and the servers.)

Other testing that has been done (most noteably by Locoweed) allowed 3-4 players to connect and actually move around in a very simplistic zone but they got such terrible lag even after months and months of "tweaking" it was often less than 1 frame per second. I know this because I took part in some of locoweeds testing and even after I abandoned 3DGS for a more flexible solution for MMO development I continued to follow his testing as I am always curious if anyone will ever get it to work.

The bottom line is this. 3DGS has a working network layer.. If you are designing a card game or word game or perhaps even a turn based application that does not require real time updating to many concurrently connected clients it works quite well. As soon as you attempt to introduce real time 3D object manipulation across multiple clients you get bottlenecked by the network code and since 3DGS does not allow owners to modify the source you are very limited in how you can handle these hurdles..

Beyond all of that, the very nature of how 3DGS connects clients makes it virtually impossible for linked servers to hand off client objects to each other.

I think if people were more open about the current potential and limitations of the 3DGS networking code instead of being so defensive it might foster a better community effort to make it better. After being ridiculed by both the admins here and some of the "preferred" users over my finding regarding the network layer of 3DGS I came to my own conclusions and found a product that works much better for my application with a much more supportive community and customer support options. If you have any further questions about my work with 3DGS and you are serious in your efforts (I dont have much in the way of free time but I am always glad to help someone avoid the pitfalls that I've already been through) please feel free to contact me on my project forums. http://www.aakrana.com/community

I wish you all the best.

Mark

Really fastlane? What kinds of interactions? I haven't seen anything other than your server accepting client connect requests?

Prence's request was regarding using 3DGS's native network layer for an MMO. I figure you have the most experience and time invested in 3DGS so would you say you have a 3DGS server and client that can interact with the level of reliability and performance required in an MMO? And if so congratulations I wouldve thought there wouldve been more hooplah about it if you had since you would be the first person in the history of 3DGS to do it.

So my question to you is quite simply. Can you show a screen shot of a simple area with 20 people walking around and interacting in it?

Mark: Fstlane has posted a video some time ago about his project. A learning simulation, multiplayer, the way many students can interact with the world and study physics, etc. It has proven to work and even physics and all the things worked on all the clients, seamless and without any bigger lag.

Just because YOU can't do something, it doesnt mean that others can't. I have been working on a multiplayer shooter project called Prototype some time ago. Basically I didn't quit doing it because of 3DGS' incapability, but more because there wasnt much more than code. My so-called partners left me alone, and they were responsible for modelling and levelling. Code-wise everything works, and even an internal test proved that we do not have much lag. It works just as fine as any professional engine(as Unreal for example). IF you actually KNOW how to use it right and how to do things right. YOu have to understand the matter you are working in before you can really claim if it is good or not.

Yes, there were many prorotypes of Prototye(strange sentence huh?) and I had big problems getting the lag away(even in LAN!) but with every time i did understand it better and better, the lag has gone away and my multiplayer moevemtn code(even shooting using trace, which does mean the you could use a MG ingame without wasting bandwidth for transmitting the position of a dummy model that moves along) proved well working, stable and lag-free even withmutliple players over INTERNET.

You have to know what you are doing.

Oh in case you dont belive me, you can download the full source(code, levels, models, anything) in the USer ocntributions forum, as i providided it to the community to continue working on it or just using the code i have done.

Yes it is dirty, but this was simply a project to learn about 3DGS' multiplayer capabilities as many people were claiming that it is just *crap*, just like you did. But it did result in that it is capable of MANY things, using the native provided 3 basic functions: send_var, send_string and send_skill. YOu do not need anything more, besides od knowledge, and time to learn the functionality of 3DGS.

As for, try to run the source of my game, try it over internet with your friends. And then tell me again if it is possible or not

@Michael: I have been an owner of 3DGS since A4, I am hardly talking through my hat. I simply posted here because Prence of Darkness asked the same questions I was asking 3 years ago. He is getting, in effect, the same answer I got back then. I thought I should clarify the response with some solid facts. I spent over a year and a half working with the A5/A6 network code, and if my findings were incorrect all I would ask to refute those findings is a working demo that allows 20 people to connect to a single server and run around without cratering their FPS to 1. Your demo does not allow that. Fastlanes working physics example allowing multiple people to view a physics interaction is not the same as even a handfull of people physically running around and sending their player object updates.

Anyways, as before, there has always been the same old rhetoric that it's all possible, it's just that there are no examples of it working.. Locoweed's work came as close to a success as anyones. I think he actually broke 10 players online with his last round of tests without completely killing the framerate..

The fact of the matter is this, 5 players? 10 players? 50 players? is 50 players enough to run an MMO? of course not.

All that Prence of Darkness is asking is has anyone succesfully made a working MMO base engine with 3DGS. The answer to that is no. Any MMO project that has been attempted with 3DGS has been unable to have more than a base handful of people connect and that is why they moved to third party network solutions.

Please don't worry about convincing me, I am very happy with my current MMO engine which I have load tested with well over 300 connections to a single server (335 I believe was where I started to bottom out). Now those were simple authenticated connections, actually 3D player clients in game I have only been able to get just over a hundred(thats WAN connects not LAN) but not because my server started to get upset, just because I didnt have anymore testers heh. At 118 connected players I was between 20-25% CPU useage and just over 10% network useage.

There is no out of the box MMO solution. My solution is not "out of the box". It took the combined efforts of a lot of people to tweak and fine tune the engine. I am not suggesting for a minute that 3DGS is not capable of multiplayer games, I believe a handful of people have produced 2-4 player games with 3DGS and they work fairly well on a LAN. ( I dont recall a successful 4 player 3DGS game that works over the WAN) Instead of just the usual rhetoric, if someone has a working solution maybe they could post it for Prence to incorporate. If not then it's all just talk.

All the best.
Mark

PS: My post is not meant to slam 3DGS is any way. I have no agenda to forward by posting and it is certainly not my intention to get into a he said/he said argument. I agree to disagree and if someone is successful with making a 3DGS MMO base engine I am sure discussions like this will bring them out and benefit the community. It's simply the honest exchange of where the network code is at that is all I ever wanted. Microsoft stopped working with the DX net code years ago because game developers couldnt make it work under load. It didnt have the flexibility required by current datastream levels for todays multiplayer entertainment titles. Trying to implement DX networking into an MMO is something Microsoft couldnt do with their multi-billion dollar entertainment division. To assume that an indie developer is somehow better than that is perhaps slightly unrealistic.

Sorry for the disruption I leave you all to your work.
Cheers,
Mark

I was in fastlane's public test a good, good while back and we had like 15~20 people in over the internet and there wasn't much lag to speak of. IMHO affording the hardware to do an MMO is more of an obstact than coding the software.

I NEVER INTENDED FOR THIS!

All I asked for is what is the proven max amount people have been able to accomplish in a shooter or an RPG.

@MarkMcFear Please don't let the community force you away from this post, it is important to have all points of view. I agree with you very much. I searched long and hard for a demo. I didn't find one. If anyone has a demo can you please provide us NOT with all the top secret code, but with a crappy down graded version that can support at least 10 or so people? I'm very close to spending the 800+ to upgrade to pro, but I can't do so until I'm sure that I wont be wasting my money. SOMEONE please post a working downloadable demo, please. I would also like for conitec to answer not just be but also MarkMcFear who claims to have spent 3000, which to me is no punk change.

It would be great if conitec would supply a multiplayer demo for A7.
I think they should. "Office" would work with multiple players, but what I mean is a complete demo, with interacting clients, some particle effects etc. Just to prove to people that it can be done. That would be great. Unfortunately they probably don't have the time...
maybe we should ask George Dan Pirvu (writer of the AUMs) I think he knows pretty much everything about c script, and it would be great of him to code a small-but working demo for massive multiplayer games...

I'm sorry I don't have such a demo. If I did, off course, I'd give you some of the essential code...

Micha

there are other options, if i remember ghostfacekillah was working on a networking plugin that was doing well, tho he seemed to stop working on it completely... if you contact him, im sure that money that you would use for pro could be used to help motivate him again hell, i'd learn how to use udp, and check it out if i could get $800

My response to Mark was meant to refute his incorrect statements and was not an attempt at getting into a flame war or to prove "quien es mas macho". Thus I will not rise (or lower) myself to the challenge of "prove this" or "show that". You either believe what I have done or don't... does me no never mind either way.

To answer Prence's question directly, with a single server and nothing else, you will not get more than 20 people in before even a fast computer bogs down. With a distributed server (that is to say all server-side functions are distributed across various computers), we've acheived 50 interacting connections with no loss of FPS and around 5% server resource usage. The only reason we couldn't get more is because we had 50 clients on three computers, one of which was running the server, and we couldn't load more in and got weird glitches from so many people on so few computers. If you read through our posts through the years you'll find this fact well documented: single server=about 20 clients; distributed server = much more than 20 clients.

BTW, it's unfair to pick on Locoweeds tutorial. It is just that: a tutorial, nothing more or less. It's not meant as a full feature optimzed network code much less an MMO backend. It is a way to get a newbie up to speed on how to network program with 3DGS. From there to MMO status is a long, ardous and mostly alone road since there are so few of us who have the time to pursue it.

Trust me, we are working hard at making our client public. Right now we are fighting our offices network topology which prevents us from making our servers public and thus giving a demo. But we are working towards that and once we are ready, you all will be the first to know!

I'm hoping to be on everyone side in this argument. Personally I like everyone in the community. In my opinion everyone has something important to say, especially after investing 300$ or more in either com or pro ed.

No offence to fastlane or anything but I haven't seen any evidence. Could someone link me to it? Further more, fastlane what where the stats of your server? (RAM, Connection type, Proccessor and so on).

There are no links and no public evidence atm.

I could provide you with screenshot with 10000 people and "10000 clients!" in big red numbers, but what would that prove? That I can use photoshop, that's what.

We aren't ready to open a server up for everyone to test it and thus, until that stage, it's merely my word (and others) that 3DGS can do what it can. I know all too well the downfalls of putting up a sample that is not ready for prime time having been at that stage many times before.

Remember, I have no stock-options in Conitec and thus I have no vested interest in the software aside from what it can do for me. I too was pissed at the beginning of my development at Conitecs reticence at telling us what the engine can do. But now we have more people working with it and at least 3 people that independently claim, from independent efforts and independent code, that it can do what we claim it can. If it couldn't handle 50 people, I gah-ran-f'n-tee you I would be the first to post it on every game engine board from here to mongolia...but it does... so I don't.

Yes, it still remains to be seen if one server can handle 100 much less 1000, yet all signs (right now) point to yes. Unless there is some weird DX9 Dplay juju that breaks the way things are scaling (and if there are articles on this I would love to see them!), then with our proven 50 clients and 2% resource use, I have confidence that we can acheive 500 clients and keep resource use to under 50%... and 500 is my magic number that proves that 3DGS (and our system) is "MMO ready". We'll see, but I remain optimistic as ever...

Running the 3DGS Server apps:
2x Xeon Duo Core 3.2
2x Raid-0 SATA HD
2x Gigabit NICs
2x 4 gig RAM (8 gig total)
1x 256 MB Nvidia (forget model)

Running the MySql server:
2x Xeon 2.6
2x Raid-0 SATA HD
2x 10/100 NICs
2x 1 Gig Ram (2 gig total)
1x 128 MB ATI (forget model)

what i dont understand is what the framerate has to do at all with multiplayer, you dont get slower rendering, your dead-reckoning just makes all object continue to do the last task they were doing, so why is anybody mentioning fps at all?

1 fps is not a sane statement unless something is terribly wrong with your connection code that causes your program to freeze or your program code in general. 1 fps usually has more to do with a function run amock. Use F11 key running from SED and see where the real delay is at.

WELL... I for some reason beyond my understanding trust fastlane and locoweed. Despite the fact that there is no real evidence (this might make me an idiot) i must say locoweed has written a very good multiplayer script. Fastlane I'm sure you can answer this about your project, are you using a DLL or C-script.

A few DLLs to connect to the MySql server.
Everything else is in C-script.

If you are doing anything MMO-ish, I highly recommend you get familiar with MySql since an MMO nothing more than just a huge database backend with pretty graphics frontend. MySql is free, it's relatively easy, and it's fast.

In most cases multiplayer will improve the framerate of your game. No longer runs complex A.I routines, instead, your entities receive updates from the server instead. Unless of course, your system for smoothing out server updates is insanely complex.

I wouldn't rag on locoweeds tutorial, he uses a different method that really isn't meant for FPS games. I've a full fledged multiplayer game(without the fancy lobbies and chat yet...) but 8 players, racing, and battle mode with 16 weapons 2 functions each enabled. This includes targeting, particles, ect. The reason for 8 players is due to my own limit(level design). I'd like to think that beyond this is possible. Remember, if your seeing really bad lag, emit some more "smoke" to hide it. Check your smoothing systems.

I cannot provide a demo to prove this right now, as I'm not going to share the code(much to complex/large to learn from anyways). However, it's really not that hard to do yourself(if you understand c-script, pointers, ect.), read through the years of "Multiplayer forums", and look for a western demo somewhere on the forums, and theres another begginers tut on AUM.

when i was using 3dgs multiplayer i had 4 players and 60npcs running smooth with a dsl as host. i wouldnt say its impossible to make an mmorpg type game with a6 but unless youre some godlike coder it wont be possible and secure.
i released the code from my previous multiplayerproject right now, its commented pretty good you can find it at my homepage http://www.ackbytes.de under download.

Thank you for sharing Ulf

I made an account just to say this.

I run impulse gaming wich is a mmorpg network. We run servers for things such as ragnarok online or lineage 2.

As an example ragnarok online uses 3 threads to run the game (map/character/login) and a fourth thread wich is obviously mySQL.

EVEN ON A LOW END SYSTEM! We can get 400-500 players online. This is with a P4 2.0ghz and 1gb of ram.

A decent multiplayer system with a server like fastlanes used would easily hold 5000+ players.

I hope this puts it into perspective for some people. If the server was having problems with 20 people on a high end PC such as that then I would not count on getting more than the 32 people that populace claims to support.

On the other hand if 3Dgamestudio supports c++ .dll support than there shouldnt be a problem with making your game first and then later on creating your own network code with c++ when you feel ready.

@ D1NG0:

As of yet I have not tested A6-based multiplayer game with more than 4 people, because my version doesn't support more players. Anyway; I guess if someone has problems with more than 20 players its either A6 fault or mostly due to one or more of the following reasons:

1: The game was not designed with multiplayer gaming in mind, to many resources.
2: The server has to many windows-desktop like sth running, no real server.
3: The server hardware configuration is not balanced to act as a real server
4: The server/mp-game is running on a standard DSL connection.
5: Someone is tampering/hacking your insecure multiplayer game/server

On the forum I read people wanting to setup their own website, mysql, mp server on 1 and the same computer with XP running behind a homegrown DSL connection. That will not work, for funplay with your mates yes for serious no.

Anyway I think its fun to learn from creating a multiplayer game. Doing a mmorpg is a little above my head though. I'm thinking about purchase Populace if I can get enough information on howto use it.

Max. 20 players per server is enough for what I had in mind. To bad Conitec didn't allowed 4 players for A6 Extra and at least 8 for A6 Commercial, guess that would be fair as with Pro they said one can get virtualy any number of players online.

The server will not be run on any of my computers. The people can setup their own server on their simple dsl connection, therefore I don't need more players. The only thing I need is to let people who run the game, choose to which server they like to connect. Don't think that is possible with A6 Commercial alone.

I hope A7 fixes multiplayer a bit. Im testing A6 right now and im not too sure what to think. I guess the biggest hurdle is that the trial version doesnt seem to support scripts at all.

Makes it nearly impossible to do the tutorials

The trial version doesn't support multiplayer, but it does support scripts.