Instances with less than 32 players

that would not be p2p, then. i don't have that level of detail on the implementation, do you have any source or is that just a guess?

Peer-to-peer doesn't necessarily imply a persistent mesh where everyone is continually connected to everyone else present, just that there is no central/static server for instance hosting.

There is definitely an instance host--possibly more than one, but certainly not everyone simultaneously--who acts as the server for a given instance. You can see this by comparing upstream bandwidth figures between various peers and noting that problems with the host have more of an impact than problems with non-hosting peers.

However I live out in the sticks, my ADSL speed is 80Mbps down / 20 up, I'd guess your connection bandwidth is rather higher?

I've had all sorts of connections since I started play ED, some lower than that in both upload and download. Bandwidth shouldn't be relevant, unless it's very low. The game client itself seems to cap upstream at 250Kbps, by default and I have never seen the game use more than a couple megabits in either direction, even after manipulating the cap.

Latency, jitter, and dropped packets are probably used for instance weighing, however.
 
Bandwidth shouldn't be relevant, unless it's very low. The game client itself seems to cap upstream at 250Kbps, by default and I have never seen the game use more than a couple megabits in either direction, even after manipulating the cap.

Latency, jitter, and dropped packets are probably used for instance weighing, however.

IME higher bandwidth is always better (no downside at least, in isolation). WRT latency I am physically much closer to the main Amazon AWS (In Ireland) than you I think, but you probably have a much fatter pipe to it & ISP traffic profiling may play a part too, I don't know.

I have no complaints about my ISP or internet connection but while there may be ways I could improve it (changing ISP perhaps, or better quality equipment in my home) fundamentally we live in different places with differing local infrastructure & network topography.
 
Peer-to-peer doesn't necessarily imply a persistent mesh where everyone is continually connected to everyone else present, just that there is no central/static server for instance hosting.

There is definitely an instance host--possibly more than one, but certainly not everyone simultaneously--who acts as the server for a given instance. You can see this by comparing upstream bandwidth figures between various peers and noting that problems with the host have more of an impact than problems with non-hosting peers.



I've had all sorts of connections since I started play ED, some lower than that in both upload and download. Bandwidth shouldn't be relevant, unless it's very low. The game client itself seems to cap upstream at 250Kbps, by default and I have never seen the game use more than a couple megabits in either direction, even after manipulating the cap.

Latency, jitter, and dropped packets are probably used for instance weighing, however.
P2P usually means an architecture where the server is replaced by another peer (client): Meaning the guy hosting the lobby is your "server". Their machine is the one new joiners connect to, via a server side matchmaking in the background. The host (server) doesn't "serve" much. the game is installed on all clients and all the host does is session management and a check on plausible locations and states. But since it's busy creating pretty graphics for the host while downloading erotic videos you can assume it's pretty busy and shouldn't be stressed too much.
 
IME higher bandwidth is always better (no downside at least, in isolation). WRT latency I am physically much closer to the main Amazon AWS (In Ireland) than you I think, but you probably have a much fatter pipe to it & ISP traffic profiling may play a part too, I don't know.

I have no complaints about my ISP or internet connection but while there may be ways I could improve it (changing ISP perhaps, or better quality equipment in my home) fundamentally we live in different places with differing local infrastructure & network topography.
Bandwidth is usually sufficient for gaming purpose except you might hit bottlenecks with upload speed.
 
Peer-to-peer doesn't necessarily imply a persistent mesh where everyone is continually connected to everyone else present, just that there is no central/static server for instance hosting.

yeah but the devil is in the details. i would have suspected more of a polivalent node architecture where every node just looks out for itself, broadcasts its status and listens to broadcasts from the other peers. the instance is then akin to local network and every node is then effectively a server and a client. it is a clean design for such a decentralized model except ... it doesn't scale well, of course. then again in elite the scale (expected players per instance) is already pretty low.

of course it could be anything in between.

There is definitely an instance host--possibly more than one, but certainly not everyone simultaneously--who acts as the server for a given instance. You can see this by comparing upstream bandwidth figures between various peers and noting that problems with the host have more of an impact than problems with non-hosting peers.

that's an interesting experiment, you mention it as if it has already been done?
 
that's an interesting experiment, you mention it as if it has already been done?

My statements are extrapolated from first hand experiences with the game. I've been in instances where I've seen a buggy host kill the instance for everyone, been the host of instances where everyone present started to complain when I spiked CPU load on my system by doing something demanding on it (without others being able to achieve the same effect), and have been able to compare bandwidth figures between CMDRs present and note that one of them usually saw much greater upstream reported by the client than the rest.

Get some people with solid connections together and compare bandwidth figures. The first one in an instance will usually be the host of it, and will tend to have a higher reported "Send" bandwidth than the rest. You can then have one of the non-hosts do something extremely CPU intensive, to the point where their client cannot be played smoothly and not affect anyone else, but if the host does this, the whole instance suffers. Losing the 'host' also appears results in more of a stutter for the remaining players than losing a random client peer. About the only thing that seems to be consistently hosted on individual peers are the NPCs that have them targeted.

I don't have any special technical knowledge of Frontier's implementation of their network model, but I'm fairly well convinced from what I've been able to observe and test that the game generally behaves as Navigare Necesse Est describes above.
 
I don't have any special technical knowledge of Frontier's implementation of their network model, but I'm fairly well convinced from what I've been able to observe and test that the game generally behaves as Navigare Necesse Est describes above.

it's possible. so that would not really be p2p then, just that instead of a dedicated well provisioned server machine with high cap bandwidth we have one random player's pc on a suburb level line which is downloading pornography. 😂
 
it's possible. so that would not really be p2p then, just that instead of a dedicated well provisioned server machine with high cap bandwidth we have one random player's pc on a suburb level line which is downloading pornography. 😂

I'm sure there is still direct connection between individual peers when possible (this is also demonstrable), but the host(s) does seem to be doing most of the heavy lifting.
 
Back
Top Bottom