Arma 3 with 1,000 players online .. possible?

[onlyrazor 11]

The thing is, 256 is achievable, but only if you know how to fully exploit the capabilities of the hardware you've got. Why do you think there are so few console games with a 32+ capability, such as MAG? All in all, it'd be interesting to see how large scale conflicts would go down in a 100+ player server.

[Smurf 12]

Would we (ARMA players) fill ONE of those servers? Don´t think so...

[PLRSniper 3]

A lot of posters in here assume a lot of things...

First off, the difference between an MMO and FPS is that MMO's are effectively turn based and as such doesn't rely so much on real time feedback.

Say you want to cast a spell in an MMO, it doesn't matter that much if it takes 20 ms or 1 second for the spell to take effect.

However, you fire a sniper rifle at a target 1,000 meters away at a moving target. If your latency (ping time) is around 100 ms that means the target might have already moved (server side) away from the bullet impact (client side) even if you "see" the bullet hit.

This can be somewhat remedied by lag compensation where you are "close enough" to being accurate so you are awarded the hit even if the target isn't in the position you see it being in client side. The problem with "too much" lag compensation is now other clients will get killed in seemingly impossible situations. For instance, walking round a corner thinking you are safe but getting caught by a bullet that seemingly flies round corners.

So, in a perfect world... Everybody would have a ping time of < 50 ms thus making it hardly noticeable that there is any delay at all. This is only possible when players are CLOSE to the server in question. (or at least have a nice clean line to the same)

The other difference is the amount of information shared between server and client. Where an MMO sends simple information per player like position, rotation and the very most basic health and appearance information...

A FPS game might send the same information ALONGSIDE information about each (*) bullet flying in the air. It's direction, speed and type. There are vehicles involved which each are equal to a player and then some. (turrets facing their own ways etc)

So the amount of information being passes around in a FPS game is greater than that of your average MMO. Assuming a pretty standard 24 bytes per player/server frame that is 24,000 bytes * 20 = 480,000 bytes/s at a server frame rate of 20. That alone would require a 4 MBit internet connection to receive. (Myself I am sitting on a 100 MBit/s so i wouldn't even break a sweat)

But now, all players are shooting their weapons. That's an additional ~16 bytes of information per bullet (that is, bullets seen by clients) which ups the requirement to at least 8 MBit and that is before we consider all the other events happening on the server.

So there is a large hurdle to overcome in terms of client internet connections alone for a 1,000 player server.

* - Most games doesn't send ALL bullets to the clients. For example, full auto fire is reduced to something line 2 bullets per second where the server gets information about 10 bullets per second from clients.

...

Now, the server itself needs to cope with 1,000 players doing all sorts of stuff. Consider that clients on full auto will send information about up to 10 bullets per second. That "may" be 16 bytes * 10 = 160 bytes just for the fire event. Now, multiply that by 1,000 (160,000 bytes/s, 1.28 MBit) for bullets alone. The server deals with ALL these bullets ballistics to ensure the server gets the final say in who gets hit and who doesn't. (Unlike BF3)

That is a HEAVY load to take server side. Additionally, the server now have to decide who is lagging and who isn't and when a "miss" is actually a "hit".

---------

I am not saying it is impossible to make a 1,000 player FPS game but we might want to take smaller steps instead of simply upping it to 1,000 players because "we can".

As for other games allowing 256 players etc, that boils down to "what did they drop in quality in favor of player count?"

Quality over quantity any day!

As for teamwork, it doesn't matter if you have 10 players or 10 million players. If a chain of command exists that WORKS the teamwork will work just fine. It will be a larger logistical challenge but it's a matter of effort...

[Tonci87 163]

I can´t even imagine a coop with 100 people....But I guess that would be awesome. The big deal with logistics, Medevac Arty, Air and Armoured support and Medevac and Infantry...

[onlyrazor 11]

I can´t even imagine a coop with 100 people....But I guess that would be awesome. The big deal with logistics, Medevac Arty, Air and Armoured support and Medevac and Infantry...

Finally some use for High Command

[teacup 2]

I sure miss the old IC-ArmA 120-140 player battles from the ArmA1 days. They were at times laggy and tended to crash towards the end (JIP was out of the question during the last hour) but it worked.

From my experience, the ArmA server has/had 3 problems, possibly related:

1. Poor scalability: the dedicated server usually maxes out one CPU core, and maybe uses a bit of a second one, but that's it. It doesn't spread the load out evenly or comparably among all available cores. This might be one of the bottlenecks. I guess multithreading is not one of the server's strengths.

2. Performance degradation. Put 80 players on the map, let them play for 2 hours, and the server FPS will sink from the initial 40-45 to a lowly 5-6 over time. It's still playable, but the risk of a server crash becomes very real and experienced players can sense the server struggling. Not sure what causes this, happens even when killed units and vehicles are cleaned up so as not to clutter the game over time.

3. JIP. Apparently this was fixed or improved in 1.60. It used to be a major pain in the ass for large MP games.

I guess only BIS could tell us exactly what's what, what can realistically be improved, and eventually what they consider worthy of spending some engineering muscle on. I for one always got a major kick out of these large (100+) PvP tournaments, but we might be a very small minority of their customers.

I have no doubt even FPS-s can have large crowds playing simultaneously on one server, much like MMO-s do (but real time). It's not that programmers are not capable of writing servers that span across multiple CPU cores, physical machines, network interfaces. Probably more a question of demand.

[Sealife 22]

What was the old ofp record , I was there it was organised by gotflashpoint iirc think it was about 250 , good times .

[SpectreTim 10]

1000 players may not be feasible... but just think about how crazy it would be to see an air assault with like 50 helos inserting full payloads of human players down on the battlefield

[Madus_Maximus 0]

It's technically possible, it was technically possible in OFP too, it's just massively dependent on good hardware and the net connections of every one involved. I'm not sure how well it'd play though or what kind of hardware everyone would require for good performance lol, never mind how beefy the server would have to be.

[cuel 25]

A lot of posters in here assume a lot of things...

-words-

Did you even read how Pikkoserver works?

Edit: Just check the OP. They're going to set a world record by having 1000 players online at the same time in a Unity engine (browser) game. All players will be on the same map and everyone is going to shoot at something/hit something and all of this will be updated to everyone. Check the 2nd post and read how _the tech_ pikkoserver works.

Imo the isn't about if we can get 1000 players at the same time, it's more about implementing a server techology that will drastically decrease server load if the server is using a cell server cluster.

Pikko Server uses a unique load balancing technology inspired by the cell phone masts used in cellular networks. Each cell server has an associated «mast» in the virtual world that determines which players it is responsible for.

A player is then handled by the cell server with the closest mast. As players move around in the virtual world, they are automatically and seamlessly switched and handed over between cell servers in the same way a cell phone is switched to the nearest physical mast. However, Pikko Server goes one step further by letting the virtual masts themselves move around in the virtual world according to the current game situation, in order to always yield an optimal division of the game objects to the different cell servers.

If a cell server becomes overloaded with players, its mast will reposition itself and handovers will be made to nearby servers automatically to accommodate the situation and balance the load.Because of this dynamic nature of the masts, cell servers can even be added or removed automatically by Pikko Server in runtime to fit the current load. This results in an unprecedented level of scalability that is not dependent on player actions and movement, relieving the game developer of the need to divide the virtual world into static zones or limit their gameplay, and enabling them to build truly limitless open-world environments.

Edited January 25, 2012 by cuel

[sandan 16]

1000 players would be a mess.. Something like an Armageddon scenario!

[[evo] dan 79]

Would be good for a full battalion attack on somewhere + their support units. Bought right size for a real formation + we can have helos and jets and ships manned to support them.

[mir00 10]

I believe this 1000 players feature is good for military training. Though i cant imagine a military base with a thousand computers specially for it.

[Tonci87 163]

Dan;2095664']Would be good for a full battalion attack on somewhere + their support units. Bought right size for a real formation + we can have helos and jets and ships manned to support them.

That sounds completely crazy...

I´ll be happy if I ever attend a 250 player coop (with ACE and ACRE and everything running stable)

[inlesco 234]

With such huge amounts of players [500;1000] the most impressive era to experience in a large battle would be WW2, IMO. Imagine what kind of "global" chaos in entire map night-time would bring. Not very accurately shooting weapons without all fancy stuff attached to them, usually met in the past, are the key to the best possible experience as a grunt.

While air, armor is likely a completely different story. But well.... 500+ fighters in the air attacking each other would demonstrate how beautiful virtual war can be.

[mattxr 9]

Id be happy with at most 200 Player Lagg Free ArmA Game... The reason being we had a Joint Coop and the Max players there was 120.. the server was fine for this and the map built for it but the game server strained under the weight of so many people.. although that was at a time before Arma Arrowhead and the MP has much improved since so i dont no what that is like.. but 200 max to be sure would be very promising. On a side note the mission was completed.. with a few people dropping out cos there computers couldn't handle it but overall it was super fun. We have a video or two of it somewhere :)

I could imagine a server full of publics at that rate would be chaos.. but clans joining together for joint co-ops is the way forward for huge games that arnt pvp.

Edited January 25, 2012 by MattXR

[CaptainBravo 0]

100 vs 100 PvP would be even better!! :)

[lwlooz 0]

100 vs 100 PvP with each of those humans leading some AI* would be perfect. :)

(* Tank,FireTeam,Squad,Platoon)

[Dwarden 1126]

from my experience with Planetside 1 (350 to 500 players /map)

and Mount & Blade: Warband (Mount & Musket mod) (250 players, 125vs125 pvp)

seems like 'playable' if there is enough scaled terrain and gameplay (aka not too crowded)

[onlyrazor 11]

from my experience with Planetside 1 (350 to 500 players /map)

and Mount & Blade: Warband (Mount & Musket mod) (250 players, 125vs125 pvp)

seems like 'playable' if there is enough scaled terrain and gameplay (aka not too crowded)

A-ha! That's a yes, Arma will support 200 players!

[enforcer4100 10]

I would already like 200 players lagg free games.

I have some nice coop scenarios for them that I need to create :p

[ScorpionGuard 10]

Most of you forget that the engine this is build upon was made to run simulation for a Brigade (3200 - 3900 persons) sizes unit for the military industries. So 1000 persons is below that mark. The only thing is to have BI to remove the restriketions from the software to do so. Also. The size of the maps where to reach up to 10000 sq km (110 miles by 110 miles +/-). I'm hoping that do to the overhual of the software. The map & persons size has increased.

[BobcatBob 10]

Im sure it will, but not by that much!

[4 IN 1 0]

What is the record of EVE Online for numbers of people within a single cluster? And remeber the lag of those huge battle is so high that later CCP have to come up an idea to slow down the time passing in that cluster so that the battle can actually be practical.

Edited January 28, 2012 by 4 IN 1

[Smurf 12]

yeah.... not big deal. I still can´t really believe that such a massive server would be populated (unless some kind of RPG mode was running, but then I wouldn´t be there) in a regular bases. An improvement for up to 300 ppl would be nice but I´m fine with what we have now.