[MUD-Dev] Geographical space paritioning

[Peter]

"ceo" <ceo at grexengine.com> wrote:

> They tend more often to coalesce into huge clumps in small areas
> (too big for a single server to handle) or into lots of small
> clumps

which is why it appears to be good think to solve by
cell-rebalancing. if a cell, which is handling a city, gets too high
load, it need to shrink to get less load (less players). the
surrounding cells will take the players. if it's too much even for
them, more surrounding cells will divide the space (which countains
out unlucky not-willing-to-lag players) take any space partitioning
algorithm (quadtree, r-tree, even b(sp)-tree).  they will
recursively partition the space until some condition is met (the
number of players in the cell is less than max-allowed, for example)

but such design would be truly overdesigned, as there would be too
much dynamic there and the system would be too easy to get to
turbulent (chaos) state.  we can do better.

> of the current RPG-style MMOG's at the moment), then geographic
> partioning is a dead-end. Mountains are obviously one of your best

i wondered quite a lot, when i saw that SWG is actually still using
it. had too high hopes for swg, probably. or they knew something
that we don't :)

> P.S. Dan, do you remember which book it was:

>   "There is an article about these movable cell boundaries in one
>   of the multiplayer game development books. It has some anecdotes
>   about how traumatic a scheme like this can be, and it was more
>   than enough to put me off."

i would like to hear that story!

Pietro
_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev