
Maximillian Power
Minmatar The Dark Protectorate
|
Posted - 2007.09.18 07:02:00 -
[3]
Originally by: cardGames
Quote:
If that were the case every other system and region would also be lagging at prime time which blatantly is not true.
no, no thats not true at all jita has a huge central population... 1000 poeople in 1 system on just a few servers... Most of the time in 0.0 unless people are fighting theirs not anywhere near that many all in 1 spot all at the same time
I'm sorry - but you are making huge assumptions here that are completely unfounded.
Your statement was that there is not enough bandwidth available anywhere to support 1000 people connected to a single server over the internet. I'm sorry. That is patently not true. The amount of data that eve sends back and forth is a pittance.
Eve is just as playable over dial up as it is over broadband. It just doesn't use much bandwidth.
It is far more likely, as a previous poster suggested, that collision detection code is probably a major contributor to the issue. Basically everything that moves in space in a single system has to look at its surroundings to see if it is touching anything. The more objects that are in space, the more live threads there are running on that node checking all the surrounding objects to see if there is a collision with one of the surrounding objects. I've no idea if this _is_ the cause, but its actually plausible, unlike your explanation which is, to be quite honest, ridiculous.
To get to the point - If the issue is collision detection code, then the issue will still exist if the server is "sharded" in the traditional sense. Lots of people on here assume that sharding will help, but they do so without the intimate knowledge of the architecture of the system to determine if that is really going to help the situation. I don't have the intimate knowledge to know if it will help or not, but I know enough, from observation, to be able to know that it is unlikely that "sharding" will resolve any of the issues that you see on a daily basis.
Moving to multiple servers per node, i.e. a cluster under a node rather than a cluster of nodes, might help, and I believe that this is something that is being looked at (again from previous posts) -------------------------------- So.... |