Monday, September 29, 2008

One Server To Rule Them All...

... And in the Darkness Bind Them.

I was reading a post on Massively about Eve's single "server" architecture (the quotes because its a cluster of servers really) and it got me thinking about how unique this design decision makes Eve.

Every Eve player in the world can interact with every other at a moment's notice, and if they choose to, fly together. Every event is local in a sense, every power bloc is neighbours to everyone else. Its especially relevant for Eve Bloggers as they can run into each other in space and exchange pleasantries if they choose. We all share the same space.

It also creates larger populations for markets, more potential pirates and victims, larger and grander conflicts involving thousands of characters over months, the list goes on. Having 200,000 accounts on one shard opens a lot of possibilities.

While it has its own problems, I hope that future MMOs will copy this model and it will become the norm instead of the exception.

8 comments:

  1. It really depends on the size of the MMO if that sort of thing is possible at all. Sure, at 200k users you can jam everyone on a sufficiently sized server cluster.... but you'd best not have any single area TOO popular or the lag will kill you.

    Also, at some point you WILL reach critical mass, and no amount of extra hardware will keep you up and lag free.

    It's a neat technical achievement to be sure - but is it sustainable if, for example, EVE grows by 50%? What about double?

    While it has its own problems, I hope that future MMOs will copy this model and it will become the norm instead of the exception.

    Unlikely.... they almost all are aiming sky-high, and want to have too many players to ever fit on one server farm. =)

    ReplyDelete
  2. Anonymous7:18 am

    Actually Andrew I disagree, recently they announced the changes that they instituted across the EVE cluster, they can now support many more players than was ever thought possible, are there limits? Of course doesn't mean CCP sits on it butt and does nothing, "stackless IO" allows EVE to continue to grow AND not suffer from lag.

    ReplyDelete
  3. Coincidentally, Ombey just posted about the frustrating lag in a huge (1100+) fleet battle. That's clearly an issue CCP needs to address, and let's hope a solution eventually is found. The "single server" model is absolutely essential to sustaining the vast politics, market, social networking, and other aspects of EVE that make it truly unique. It would not even be close to the same game using a different model.

    ReplyDelete
  4. Manasi:

    The limits are going to be based on how occupied a given logical section of the game is. I would assume that the best model would be to host all of the players in a certain subset of the game world on one physical server within the cluster, and as they move around, the jump from server to server.

    Architecturally speaking, this would allow the game to broadcast packets containing all pertinent data to all players on a server without having any internal network communication (or, very little).

    Now, assuming that the playerbase grows while the gameworld does not, you can handle more load by making individual servers host less and less "space" (and thus players) which will work for a time. However, in densely populated areas, you then start to see internal server-to-server traffic increase, which introduces lag of a different type.

    All in all it's a pretty cool architectural problem to try to solve, and I commend EVE for getting so far with it. However, I do believe that if playerbase growth exceeds the capacity of the physical servers, or there are areas that are too populated (like in the battle mynxee references) then the model is susceptible to huge lag.

    A game with a player base of half a million would be hard pressed to host all of the players in a single game world.... and let's not even try to think about something like WOW. (Even ignoring the technical boundries, 10 million players, or whatever, jamming the landscape would be insane.)

    ReplyDelete
  5. Personally I think WoW is an outlier example in that MMOs of the future will be considered successful at the 500K subscriber mark, and beyond expectations when approaching 1 mil. I just feel that the future hold more of "many smaller MMOs" rather than "fewer HUGE MMOs".

    With that in mind, can an MMO like Eve with 500K subscribers survive on one server cluster. I think (and hope!) that network load-sharing and communication technology will be able to address the lag issues going forward. I have no solid evidence to back it up but CCPs recent innovation of StacklessIO that allowed the black hole of the Jita system to become unusaable at 500 players to completely smooth at 1000 players holds great promise for the future.

    Time will tell :)

    ReplyDelete
  6. Personally I think WoW is an outlier example in that MMOs of the future will be considered successful at the 500K subscriber mark, and beyond expectations when approaching 1 mil. I just feel that the future hold more of "many smaller MMOs" rather than "fewer HUGE MMOs".

    That WOW is an outlier is not a point of debate - I think that everyone will agree that it is. Your view of "the future" mirrors the current reality, but that doesn't change the fact that EVE's goal of hosting all players on a single server is, and will continue to be, an aberration.

    My point is that almost all new MMOs come out trying to be a WOW-killer, and thus trying to claim as much of its playerbase as possible. Because they believe that they have a good shot at this (or are going to go down trying!), they implicitly support an architecture that is based on fragmenting the playerbase into bite-sized chunks. The driving reasons here are:

    a) It's easier (and thus cheaper) to design and support
    b) What big name publisher is going to support development of a game that is only targeting a small audience? The motivation is to make as much money as possible - and WOW is the model for how this is accomplished.
    c) All developers dream big - why aim for mediocrity?

    ReplyDelete
  7. While it may be easier and cheaper to design and support, I think that the single server architecture will be more attractive as the technology and methodologies propogate, hence to be competitive and unique with a more immersive gaming experience, developers will push the single server model to attract the increasingly demanding and diffuse MMO dollars.

    There is so much more that can be done with tens of thousands of players interacting as opposed to only thousands. Its the future and Eve is merely the pioneer.

    ReplyDelete
  8. I continue to disagree, as you well know.

    I ran the numbers earlier today. Assuming 10m subscribers and 440 realms, a typical WoW server (realm) hosts AT LEAST 22,000 accounts. This, I think you will agree, is far more than "a few thousand" players available on each server.

    Now, some realms are sparse, while others are crowded so the numbers probably vary a lot. Also, many players have accounts on multiple servers, so the number of server-accounts exceeds 10m, but who knows by how many.

    Also, I've posted more in your other thread - but I think that after a point you reach a non-technical boundary where there are TOO MANY PLAYERS on the server, and the game experience starts to suffer even if the hardware hums right along. Humans can only take in so many things at once.

    Another thing to consider is that all of these multi-server games are tied together by a community that eclipses the actual server they are on. The WoW-blogging explosion is an example of that, but also Blizzard has their forums, etc. This is not unique to WoW, BTW, you'll find it elsewhere. Whether or not I can actually run with 100% of the 10m subscribers (or 500k for WAR, or whatever) is not that big of a deal.

    ReplyDelete