Didn't see this posted, but since its a widely discussed topic, I figured I'd post this:
http://www.in.techradar.com/news/mi...-relearn-how-to-play/articleshow/56960711.cms
"There's been a lot of discussion around dedicated servers vs P2P. Can you talk through the decision there?
JV: One thing that is not entirely clear to people is that we don't actually have a traditional peer to peer architecture. We have a new type of architecture that, while it's sort of based on a peer to peer philosophy, is actually there in order for us to do this game with eight players active simultaneously and then these two hundred AI running at the same time.
We were thinking about what to call it, but it isn't actually traditional peer to peer.
A lot of the complaints that I'm seeing online, not all of them but a lot of them, are actually just lag issues that have to do with normal network states.
That said, the game is built on that framework because without that framework there wouldn't be a game. We wouldn't be running the way that we are right now. I wish I could go into more detail.
At this point, it seems to be working the way that we anticipated. I'm not terribly concerned about the problem right now, but we're going to continue to look at the needs of the player base as we move forward.
I think there's often a sense within the community to read something off a spec sheet and then decide based on that whether something is good or bad, but then the implementation of that spec is often the far more important aspect of it. It sounds like your implementation of P2P ultimately is one that you're happy with, even if on the sheet it technically says P2P.
JV: Yeah, we run on a thing called ‘the simulation'. The way that our network system is built is that it's built around everything being 100% fair. So what you're seeing is what I'm seeing at the same time, there's no visual advantage in the game.
There's no host advantage?
JV: There's no host advantage. This is the part that is not well understood because our technology has not been used in the wild before on other games. We've been using it all along, but it's not something the audience is familiar with.
In our peer to peer solution it's not really peer to peer, it's based on that.
There was this moment that was really important. Six or eight months into development I had this moment where these engineers came to me and said, ”So, you want to do eight players running over a network and you wan to do that with two hundred AI, over a normal internet connection? This is your pitch?"
I said, ”Yes."
They say, ”That's impossible. I don't know if you understand how networks work designer boy but that can't be done."
Fortunately I work with these incredibly bright bulbs...and they come back a couple of months later with this thing called a simulation and they took a bunch of white papers for Siggraph which they assembled it into this new way of doing networking code, but when communicating it as a peer to peer model the problems with the other implementations kind of scare people, which is perfectly reasonable.
So we'll just see moving forward how it goes, and if there are actual problems then we'll reassess and see what we need to do."
http://www.in.techradar.com/news/mi...-relearn-how-to-play/articleshow/56960711.cms
"There's been a lot of discussion around dedicated servers vs P2P. Can you talk through the decision there?
JV: One thing that is not entirely clear to people is that we don't actually have a traditional peer to peer architecture. We have a new type of architecture that, while it's sort of based on a peer to peer philosophy, is actually there in order for us to do this game with eight players active simultaneously and then these two hundred AI running at the same time.
We were thinking about what to call it, but it isn't actually traditional peer to peer.
A lot of the complaints that I'm seeing online, not all of them but a lot of them, are actually just lag issues that have to do with normal network states.
That said, the game is built on that framework because without that framework there wouldn't be a game. We wouldn't be running the way that we are right now. I wish I could go into more detail.
At this point, it seems to be working the way that we anticipated. I'm not terribly concerned about the problem right now, but we're going to continue to look at the needs of the player base as we move forward.
I think there's often a sense within the community to read something off a spec sheet and then decide based on that whether something is good or bad, but then the implementation of that spec is often the far more important aspect of it. It sounds like your implementation of P2P ultimately is one that you're happy with, even if on the sheet it technically says P2P.
JV: Yeah, we run on a thing called ‘the simulation'. The way that our network system is built is that it's built around everything being 100% fair. So what you're seeing is what I'm seeing at the same time, there's no visual advantage in the game.
There's no host advantage?
JV: There's no host advantage. This is the part that is not well understood because our technology has not been used in the wild before on other games. We've been using it all along, but it's not something the audience is familiar with.
In our peer to peer solution it's not really peer to peer, it's based on that.
There was this moment that was really important. Six or eight months into development I had this moment where these engineers came to me and said, ”So, you want to do eight players running over a network and you wan to do that with two hundred AI, over a normal internet connection? This is your pitch?"
I said, ”Yes."
They say, ”That's impossible. I don't know if you understand how networks work designer boy but that can't be done."
Fortunately I work with these incredibly bright bulbs...and they come back a couple of months later with this thing called a simulation and they took a bunch of white papers for Siggraph which they assembled it into this new way of doing networking code, but when communicating it as a peer to peer model the problems with the other implementations kind of scare people, which is perfectly reasonable.
So we'll just see moving forward how it goes, and if there are actual problems then we'll reassess and see what we need to do."
