On the Feasibility of Single-Server MMOs 316
GameSetWatch takes a look at the issues involved in creating an MMO that does not split its users among many different servers. They suggest that running a single "shard" is the next step in the evolution of MMOs, since it better allows player choices to have a meaningful impact on the game world; supporting different outcomes across multiple shards is a technical nightmare. They estimate, from the hip, that the cost to develop the technology required to support a massive amount of players (i.e. far more than EVE Online) on a single server to be roughly $100 million. Another recommendation is the strong reliance on procedural and user-generated content creation to fill a necessarily enormous game world.
Re:Lag. (Score:5, Interesting)
"Half Sharding" like Guildwars has solves that.
Is town too busy? Boom, new set of districts is spawned! They will be probably located on different server too, making players overcrowding non issue. Players can switch them at will (as bonus, they are grouped by geolocation and laguange, but can still switch to different ones.
The only way to handle rush of thousands of players to one area when special events happen. 5 thousands of players want to participate in xmass feestival? no problem, just spawn 50 districts in that town.
Oops (Score:2, Interesting)
Re:Impact? (Score:5, Interesting)
EVE is a very good example of the interest of having a single universe where players actions do have an influence, but it is not for everyone, some people will get intimidated by it. However, I don't understand why they think that the EVE model won't scale. They already have 300 000 players and their system is quite simple : different stellar systems can run on different servers. You cannot interact with something that is not in the same solar system, for that you have to "jump". It is easy to add more systems by adding more servers. It does scale up.
Actually, I see an even bigger problem (Score:5, Interesting)
Actually, I see another even bigger problem, at least for more traditional (WoW-type) MMOs. How big should your world be?
Too little player density => people start complaining that it's pretty much a single-player game like Oblivion, except you occasionally see another player. Many games ended merging up shards more for that sensation of empty space than because of costs. (It's equally easy to just merge the physical servers inside a shard, to support a lower population per shard, if you're only concerned about hardware costs.)
Too many players on too little surface => lag (think: landing in Ironforge, back when it had the only auction house for Alliance), routinely having 5 players camping the same mob, and generally it just starts feeling cramped. Again, you have players starting to complain.
Basically if you want to be single-shard, you have to essentially guess how much population you'll get. Maybe just within the right order of magnitude, but guess nevertheless. It's not that trivial. On one side of the guessed-wrong spectrum you have WoW which got launched with only a handful of servers and had massive queues, on the other end of the spectrum you have more than one game who thought they'll be teh WoW-killer and then had to merge 4 servers in 1.
Merging or splitting shards is an easier way to deal with that problem than having to physically add or remove new areas, to fit the population.
Additionally, world size influences other things, like travel times, exploration, etc. There is an ideal apparent size where people don't feel like they're being packed like sardines and running around a back yard, but don't go "fuck it, I'm not spending another hour just running back to the quest giver" all the time either. It's easier to fine tune that if it's its own problem, orthogonal to everything else, than when it also has to fit the population numbers.
Basically if EVE's game type was well suited for that kind of one-shard world, more power to them, but for other types of MMOs it might actually be a bad idea.
Re:Lag. (Score:5, Interesting)
Sharding's just the easiest solution, and hence most common. It's really not necessary any more, and detracts a lot from the social side of sharded MMO games.
Re:Lag. (Score:4, Interesting)
Well, you can easily solve this, if you allow some more strong MIP mapping technologies for everything, including the transferred packets, models, etc.
For example you could just transmit the data of the most relevant people and objects, until the pipe is X percent full, then use some "group" model, with only simple coordinates and very simple models etc for every player outside of that relevance radius. And so on... in a curve that is somewhat between quadratic and cubic (depending on the dimensions of your relevance space).
Do not forget to include the main viewing targets of your player into the calculation, so they can still target something far away, and see it in a good quality.
You could even make it work like interleaved JPEGs, where you only load the roughest details at first, and then become more and more detailed, the more you want to view it (=the more you wait).
I think people would prefer that to having every single object that is visible at all to be transferred and rendered in some kind. Nobody cares about the quality of stuff he does not care about. ^^
Re:Lag. (Score:5, Interesting)
In EVE a solar system is a discrete "zone", there are many thousands of solar systems. Each one is assigned a node on the server (blades actually) and that node may and probably does host more than one solar system. They have had limited success in beefing up big fights by moving solar systems with expected fights to a node by itself. Thereby offering as much power as their architecture will allow. However, they can only do this at "down time" which is one hour daily, so if a big fight breaks out in the middle of the day, there is no chance of it going well. Most of the REALLY big fights happen in systems that often have very low average traffic, so they are assigned to a shared node. Then for 8 hours, that system has 400 people in it, and the node is past it's limits very quickly. If they had coded the game so that a single solar system could use more than one node at a time, they could brute force the problem away entirely. But that isn't possible the way it's built.
Even so, the EVE cluster is/was on the top 500 list of super computers. You can't say it's not for lack of trying.
Why yes, I WAS an EVE player. From Beta till about a year ago. I finally gave up after countless fleet encounters were destroyed by CCP's clever, but impotent load balancing. The breaking point was when I realized that even when we had a dedicated node for every solar system in our territory, we still couldn't have a full out fleet battle without crashing the node. I'd have been happy to get half our fleet into combat, but we couldn't even do that. Granted, we had 800 ships or so and our opposition had at least 1000. I've yet to see any game that can put nearly 2000 players on a battlefield and still function.
CCP does get credit though for effort. 3 years ago you'd be LUCKY to pull off a 200 man fight. Now you can put 500 or so into a system and get your fight on without major game breaking things happening. It won't be silky smooth, but you can get it done. Ironically, 3 years ago a 200 ship fleet fight was a rare and wondrous spectacle. While a year ago, I could assemble a 200 pilot fleet in 20 minutes. So what was a major event is now a typical saturday night. The servers got better, a lot better, but they aren't keeping up with the players.
Re:Lag. (Score:3, Interesting)
From a packet lag (rtt) point of view, does physical geography enter into it at all? I see 200ms ping times once my packets start getting around to the other side of the world...
Re:Actually, I see an even bigger problem (Score:3, Interesting)
The problem with a single shard and part of the reason I gave up on EVE was that everybody playing the game has the same environment. I was looking for more roleplay out of the game but although there are a few corps playing the game in that manner, none of the people you meet in space are.
I was stuck trying to maintain a consistent character story while everyone around me was talking about things in the news, who got fired at work (IRL) etc. I wouldn't attack another player unless it was inline with the character but that doesn't stop everyone else attacking me without any pretence. Towards the end I even started playing by my own death rules, if the character dies, thats it, he's dead, he and all his assets are destroyed and I start again but that just made me a target
I was praying that EVE would split into different shards for different game styles, even if that meant the RP shard was largely empty of players (which would have been a draw for some players anyway)
Re:Impact? (Score:2, Interesting)
Re:Completely the wrong approach (Score:3, Interesting)
I thought of a solution to the "game pausing" problem.
When you organize into a team for the "month long game of Warcraft 3", you choose playing hours. So, you might "sign up" for a game where everyone will play from 6pm to 10pm each day, central standard time. Or any arbitrary set of times, obviously this would mean that most of the time, you'd be playing with other people from your time zone.
This would make the game more intense : every second actually playing the game would be precious, because if you just stand around, other players on the other teams would be beating you to the quests. Obviously, there would be some randomization : the statistics on weapons and abilities would change randomly a little bit with each match, and the locations of the dungeons and stuff would also change. It would also be possible to play on a "map" that was custom made by other players.
And if you're a casual player, you could sign up for a "3 days a week, 1 hour" match where you only play a total of 3 hours a week. None of your competitors would have any more time than you, or be ahead at the from the start.
As for the 1 month expiration date on your powers : have you ever noticed how the most fun part of any new MMORPG is that level up during the newbie levels? The game stops being as much fun once it starts taking forever to level up. And you can't PK until you finish the chore of getting to the top level.
In my vision of a game, you could obviously go try to gank someone on the other team at level 1.
Re:Lag. (Score:2, Interesting)
Example [youtube.com] of a fairly decent siege, even though it's still prior to people finally unlocking the higher end siege weapons which are just now starting to appear on the battlefield.
I can't wait to pilot a Man Of War!
Re:Lag. (Score:5, Interesting)
"I've yet to see any game that can put nearly 2000 players on a battlefield and still function."
Then you have to look for games designed from the GROUND up to support such play. Typical 'sword and sorcery MMOs' that are built for perhaps up to a 'raid' of 50 players are simply not going to scale (without ridiculous hardware requirements) to 1000 players impact-free.
OTOH - there ARE games that were built to simulate the actions of hundreds of players in a single area, so scaling to thousands is logically a lower burden. www.battlegroundeurope.com (formerly ww2online.com) catches a lot of crap for a graphics standard that doesn't approach the bells & whistles you see on today's (shoebox-arena) shooter, but I've played in many battles that (in a multi-square km area) with at least a thousand players including multi-crewed vehicles, hundreds of infantry, and a swirling aircraft battle overhead fighting for hours for a strongly-defended city. It doesn't happen often (playerbase is now only about 12k) but I just use it as an example of a system that WAS designed to handle such a load.