Curse

scytherage · Aug 07, 2005, 05:52 AM // 05:52

I like this idea so much, and I would love to see it implemented in some way in the game! Check it out:

HOW TO IMPLEMENT LARGE SCALE BATTLES IN GUILD WARS

Here's a suggestion for arena net--if it isnt possible to go beyond 8 players per team, why not create some kind of game type where guilds can enter in teams of 8 and battle simultaneously in instanced areas?

In the game type that I am envisioning, there are several 8 man teams in a guild that 'enter a mission'. Another competing guild should have enough players to compete in this type of epic battle. So if there are 8 teams of 8, there are 64 players competing in 8 different instances, wherein the result in each instance will directly affect the battle in other instances. Each 8 man team has a commander who can communicate with whoever is the 'warmaster' of the entire guild at the time of battle. Each 8 man team is in a different 'area' of the entire battlefield (and that area is essentially one instance), and each 8 man team has a goal in order to defeat the other 8 man teams and eventually defeat the other guild entirely. For example, if the battlefield is a castle, you have the castle front, the castle rear, the inside of the castle, and so on....each area/instance has 16 players fighting it out, and whenever one team wins, that will affect all the other instances in some way.

For example, one 8 man team has to attack the rear of a castle, while another 8 man team has to attack the front. The other guild has teams that should counter these two teams. If the 8 man team at the rear destroys the rear of the castle, they enter another instanced area and proceed to killing the guild lord, while the other 8 man teams are fighting the guild in other areas of the whole castle. For instance, if one 8 man team manages to launch a catapult, this would affect the battle inside the castle and destroy its wall (in that case, one instance will 'tell' the other instance, team B has launched catapult, destroy wall!) and this would further damage the castle occupants. On the defensive side, the 8 man team defending the castle can be defending some conjurers who are trying to spawn monsters in order to make the battle more difficult for the 8 man team that is trying to kill the guild lord. So at the last minute, the 8 man team in the castle can get reinforcements through an event that is happening in another instance.

That way, you can have epic 16-32-64 player battles without losing that perfect number of 8 players per team, so that lag doesnt become too much of an issue, and the network is best utilized (less cost for arena net because each 'area' of the battle field is instantiated).

KelvinC · Aug 07, 2005, 07:17 AM // 07:17

/signed

Thomasuwoo · Aug 07, 2005, 09:01 AM // 09:01

I like the idea. I mentioned this in an earlier thread of mine here. It got lost under alot of other threads. While not a detailed as yours it offers a differnt approach. Why not combonie the two ideas together? Actually to save time I'll just quote my post here.

Quote:

Originally Posted by Thomasuwoo

Not what it looks like.

I've been thinking about the problems that europe has been having as of late and it got me thinking. The reason Alot of people are leaving Europe is not just because of the lag (Which I've heard is horrendous) but because their region rarely has favor and thus they are missing out of alot of game content.

What I thought would be a great Idea would be an area similar to the temple of ages (Ie pay to get in, tough as nails, nice rewards) but with a catch. It could only be accesable VIA the international district and can only be accesed if each team has at least one member from every region.

I'm not sure on what details this dungeon could have but the point is that I bleive the regions need a little bit of co-operation instead of being at each others throats all the time. Somethimes I think Guild Wars might lead to WW3 with the tension between nations!

A few ideas though, rather then just a simple dengeon wouldn't it be great to throw 3 parties from each region into a PvE area and make them work together? Maybe at first putting them in differnt loactions, randevousing and then trying to move onto the UBER areas?

I do realise the problems that can arise from such a setup but surley it would be worth it. I'm all for competative play but the gulf between regions is so wide it scares me.

Any further Ideas/Suggestions?

This way we're keeping the epic battles going but combining them with otehr regions. Maybe too many steps at once?

Either way i like your idea and it has alot of potential. Also rember that the burial mounds in the tombs have 64 players in them. Having 3 or so teams fighting small (But powerful) groups of monsters should still be able to be archived without putting too much stress on the server.

scytherage · Aug 07, 2005, 09:25 AM // 09:25

Hey, that's a great idea! I always thought of my idea in terms of PvP; but in terms of PvE it can also work! I like it

In PvE it can be used to tell a story on a grand scale, where certain characters/groups can be heroes when they accomplish goals.

Since arena net is already working on observer mode, they can even add it to this idea, so that if a team already finishes their goal, they can watch the other teams of the guild as they are accomplishing their goals as well (whether it be PvP or PvE).

*Prays to the gods at Arena Net to consider*

scytherage · Aug 08, 2005, 10:37 AM // 10:37

I just realized that implementing this idea for PvE would also be good because this would remove the need for having an opposing guild of the same size when undertaking some kind of large battle...so every type of gamer (PvP, PvE) can get to enjoy a large scale event with less hassle.

strcpy · Aug 08, 2005, 12:55 PM // 12:55

The problem is that this will not scale well. It would be easier to have a one large arena. At least in that case much of the information can be multicasted effeciently to the clients and all the instanced controllers would automatically be on the same machine, or at least load balanced more effeciently.

The whole reason that instanced worlds scale very well on server load is that each instance does not have to communicate with any one else (along with few, if any, global persistent states that can be twiddled by each player) - you are pretty much at a O(n) growth, with most of the computation done on the client end (by Amdahl's law the greater the part that can be parallelised the closer to a speed of 1 you get).

Now, with your set up status from each instanced region must be co-ordinated with each region. To have this work you will first need to make sure that every thing happens in a timely manner and that it happens in the correct order - as the meta-maps grow larger this gots to actually be a difficult problem. Even if we assume that players will not be irritated that the "effects" aren't propagated in a timely and correct fashion (say managing that the "outer wall" team won giving the "inner court yard" a bonus, but the inner court yard loosing before it is applied and distributing that effect outwars when it shouldn't) then you have a O(2^n) growth - and that among several different instanced worlds and having to always be done at server level (meaning high server load).

Could Anet do it? I don't know - don't really know thier server layout and how much of an interprocess load they can handle. It begins to matter of you are running on large shared memory SMM's or large cluster farms, what your interconnects are, how load is balanced across machines. On a large shared memory SMM a 2^n algorithm with small instances of n could ne pretty easy to handle - but that's why those machines cost millions of dollars (and what they are designed for). Under the more typical cluster farm that scenario is highly inefecient.

Though you could get this to somewhat work in PvE - the main difference being updates. For PvP it is going to have to be timely - it doesn't do anygood for the statuses to be accumulated in a global counter and applied every ten minutes to any new instances of a game world (this would be similar to the favor of the gods).

There is a HUGE difference in performace cost (and the algorithms to provide it) between having something *now* and having it *soon*. PvP requires now, PvE requires soon.

scytherage · Aug 08, 2005, 03:12 PM // 15:12

I'd like to say that I dont admit to understanding Anet's server architechture either, I simply thought of it as, when an event happens in one instance, send message to other instances. If there are four other instances, each possibly hosted by one machine in the group which acts as a server, Send message to these machines. Then those machines distribute that message to all the clients.The bandwidth load in this case (at least how i imagined it) would no longer be in anet's hands, it would be in the 'server' for the 16 players that are in a certain instance, so the bandwidth load is lessened for the main machine(s) at anet. Anet machine sends to five other machines, and that's it. Five other machines send to their clients, and that's it. Managing and synchronizing everything can be handled by prediction and whatever other magical ingredients are already in the programming....Five machines receive OK from their clients then send the result to ANET...I dunno, i really just simplified the idea. My basis for the idea, really, is my own personal guess that one of the reasons that the game is currently free is that somehow the whole concept of 'instancing' areas has lessened the bandwidth cost for arenanet, and if this is used in some other creative way, Anet can come up with a whole new gametype without losing whatever design they created for their network communication.

Of course if syncronization is an issue, it can always be done that the battles will not all be simultaneous. It can be two groups of 16 first, followed by another 16....the implementation of my suggestion is only limited by the creativity of the game designers. The point is, to have some kind of larger scale battles that takes advantage of whatever infrastructure Anet has created which has somehow allowed their service to be free.

Anyway, it's good to hear a technical idea of how it 'might' work....thanks.

KelvinC · Aug 08, 2005, 03:38 PM // 15:38

Quote:

Originally Posted by strcpy

Now, with your set up status from each instanced region must be co-ordinated with each region. To have this work you will first need to make sure that every thing happens in a timely manner and that it happens in the correct order - as the meta-maps grow larger this gots to actually be a difficult problem. Even if we assume that players will not be irritated that the "effects" aren't propagated in a timely and correct fashion (say managing that the "outer wall" team won giving the "inner court yard" a bonus, but the inner court yard loosing before it is applied and distributing that effect outwars when it shouldn't) then you have a O(2^n) growth - and that among several different instanced worlds and having to always be done at server level (meaning high server load).

You assumed all the instance inter-communication is handled by the server. Actually it can be implemented as an instance over instance communication model (the host of the each inner instance acts as a peer to the outer instance). So it will be O(n) instead of O(2^n) due to high parallism (host is 2n and rest is n). The server only need the end result (win/lose?) and this is basically negatable.

For the synchronization, multicast with time-stamping will do (it is no different than the existing instancing).

scytherage · Aug 09, 2005, 12:04 AM // 00:04

Quote:

Originally Posted by KelvinC

You assumed all the instance inter-communication is handled by the server. Actually it can be implemented as an instance over instance communication model (the host of the each inner instance acts as a peer to the outer instance). So it will be O(n) instead of O(2^n) due to high parallism (host is 2n and rest is n). The server only need the end result (win/lose?) and this is basically negatable.

For the synchronization, multicast with time-stamping will do (it is no different than the existing instancing).

Yes, my idea is something like that. One of the machines in each 16 player group can act as its 'server' and all it has to do is send the result of the match to the other instances.....

As long as not all communication is handled by a single server, as is the setup in most massive battles, this type of idea is feasible.

strcpy · Aug 09, 2005, 10:08 AM // 10:08

Quote:

Originally Posted by KelvinC

You assumed all the instance inter-communication is handled by the server. Actually it can be implemented as an instance over instance communication model (the host of the each inner instance acts as a peer to the outer instance). So it will be O(n) instead of O(2^n) due to high parallism (host is 2n and rest is n). The server only need the end result (win/lose?) and this is basically negatable.

For the synchronization, multicast with time-stamping will do (it is no different than the existing instancing).

At least, as far as I can tell, the server is God and makes the ultimate decision (and I believe that was thier statement also) - thus it needs more than "win/loose". The above scenario is ripe for hacks and is why Anet avoided such a thing. If the above were to be true then I don't think my char would be randomly warping around today (my upstream maxed with bittorrents and the server never knowing I changed direction or stopped running - quite irritating as I was playing a solo monk and this aggro'd too many and I died several times).

If you want to accept local hacks, then I agree - unless the ultimate decision is on Anets server it's a hack envoronment (and if Anet's is the final arbiter it has to know everything anyway, no real gain to client sharing). If not, and the server is God, then my analysis is correct (the other option is that if there is a descrepancy one could call the whole match a "tilt" in pinball speak - but that would also hurt legitimate network issues and msot liekly be common). I suppose the acceptance or not is dependant on what effects you want.

That can be a hard line to choose from - server law or client law. Instanced regions with server being God works pretty well as far as scalability goes - pretty much O(n). When instances start having to communicate and those communications have a real impact then it becomes more hard. Easier to do limited large battles. There are other solutions, you could have a group barrier to entry (win so many times in a PvP battle before allowing entry) and that would limit n. It may be that an "acceptable" comprimise is made - hacks that aren't so much that the player fuss, load such that Anet doesn't care - though I really doubt it.

And to note, I figured there would be a complaint that it isn't strictly O(2^n) - each battle is O(2^n) and the battles are O(n) - I'm just too lazy to figure exactly what it would be.

Ultimately you are either in controll of the game - in which you need to know pretty much everything and see if it is sane, or you allow the clients most control - in which case you get hacked like mad (even if you do a round robin or random control - the hacker will eventually have an instance in thier controll and take full advantage of it).

KelvinC · Aug 09, 2005, 11:29 AM // 11:29

Actually the inter-communication between instances don't need to be hard-real-time. A soft real-time would work. inter-instance event can happen for each 5 seconds period is good enough.

The inter-instance events are usually like the gate is down, or something is captured/destroyed. 5 Seconds delay isn't that noticable.

scytherage · Aug 09, 2005, 01:31 PM // 13:31

Here's something to chew on: Arena Net has actually already considered this idea in the past, and they are looking into it, among many other things that they want to add to the game... Thanks to CodeOmega from gwvault.ign.com message boards for the information. Here's what he said:

Quote:

Originally Posted by CodeOmega from gwvault.ign.com

I posted this the last time someone inquired about large scale combat. I figured it's worth a repost.

Taken from Fansite Friday #40, Arpil 8, 2005 (http://www.anzgw.com):

"Are there any plans for cooperative PvPs / PvE where several teams can group together and fight large scale battles versus other teams of players or groups of "epic?" monsters. Are there any plans for the ability to play in a relatively larger battle that involves several teams working together trying to achieve the same objective on the same map?

At this time, the focus of Guild Wars is on battles that involve direct head-to-head combat on a level that allows a great deal of strategic depth. As you know, we believe playing in a team of eight players offers you the best opportunity to develop and use complex and refined gameplay strategies. We don’t support 50- or 100-member warring parties, because for the most part, a whole lot of strategy and skill flies out the window at high numbers. The loss of those key game elements would be one reason why we might hesitate to support allied team play that involves a large number of players.

I’ve taken part in large-scale combat, and I have to say it was so un-fun that I left before the battle was finished. Now I’ll be the first to admit that this is just one player’s opinion, and that I didn’t care for it because the sort of hectic, disorganized onslaught combat I experienced just wasn’t my cup of tea. It might appeal to any number of other players, though.

I guess what I want to say is that as a player, I believe that large-group combat can surely be better conceptualized than what I’ve experienced. If the Guild Wars design team takes on the challenge of allowing wide-scale team alliances, I feel sure that the outcome and the experience will be altogether superior to the large-scale battles players have been offered in other games. After all, Guild Wars is and will remain a game that is centered on player skill. So in designing for this type of large-scale combat and for team alliances, the Guild Wars team will be careful to assure that the promise of rewarding players for their skill is upheld, even when there are a large number of players involved."

-Gaile Gray

My translation of bolded text: Arena Net has already thought of this idea and think it is possible, and currently they are looking into its implementation. Whether or not we will see it in the future is not certain, but they ARE looking into it. w00t w00t!

KelvinC · Aug 10, 2005, 01:18 AM // 01:18

Quote:

I’ve taken part in large-scale combat, and I have to say it was so un-fun that I left before the battle was finished. Now I’ll be the first to admit that this is just one player’s opinion, and that I didn’t care for it because the sort of hectic, disorganized onslaught combat I experienced just wasn’t my cup of tea.

Not if it is done right.
Larger battleground need squads (sub-teams)

Rieselle · Aug 11, 2005, 02:00 PM // 14:00

We already have pve areas with independent zones, why not a "game-board" style large scale map, divided into several interconnected zones? And limit the number of teams that can enter a zone (maybe one per entrance? That way you can flank and outnumber the enemy with careful maneuvering). The objective is to "capture" and defend as many zones as you can. Think of various classic wargames which are composed of nodes connected by lines, and you move "armies" from node to node etc.

That would give even greater scope for strategy than a mere 60v60 zergfest - you'd have use real large scale strategy. And every zone is an independent instance, so very little cross-communication is needed.

And we can then have a wonderful variety of maps for each zone. Perhaps the zones at each end would be a castle for that side, we could have different sorts of terrain for the zones in the middle, with towers and things that could be used by the first team to enter against latecomers, etc.

It would feel like a proper war, with trekking with your squad through the swamp to the forest, to wait in ambush for enemy troops, suffering losses, pulling back to nearby keep held by an allied squad, sneaking around and flanking the enemy, cutting off their escape route, catching them in a pincer formation... etc, etc.

Heck, instead of lackluster pve they should have just implemented a huge pvp world for us to play in.

GhostPoet · Aug 11, 2005, 05:50 PM // 17:50

no matter what..I won't be happy with PVP until we see Fort Sieging.