Jesus H Breakdancing Presley.
Another one of these threads?
Come on guys.
Another one of these threads?
Come on guys.
Almost a year later, Multiplayer is still a mess
- Thread starter Toumal
- Start date
- Status
OP makes a valid point though.....
If you played Eve as static point and click you were doing it wrong. Using traversal velocity was key to mitigating damage while damaging. Sometimes it was best to joust if your weapon tracking was slower than your opponent but you were faster. And, there was always stasis webbifiers as well as warp inhibitors and all sorts of EM you had to know how to deal with. Complex.
And as to the universe. ED is procedurally generated but empty. Eve's galaxy may be static but the territories are dynamic because of player groups goals, war or diplomacy.
Last edited:
Those elements didn't add any complexity, still a target and click, no effort to hit targets at all, many missions ends with you just targeting and forgetting about it for a while
Well we'll just have to disagree because it seems I played Eve more actively than you did.
Now I've had time to properly digest this I am in agreement with you OP. It is possible without a huge amount of work to implement server side peer to peer hosting of an instance, this has been done in several games for tournament's already.
Numpty talk example:
Server acts as an invisible host to which other clients connect.
It's only networking, not rocket science!
Those elements are eminently complex. You can't just slap purples on things and away you go across space; EVE combat is like Vulture fitting crossed with MMO-style systems interaction. Point and click is a boring combat flight model, agreed; but the combat itself is rather deep and strategic.
1upped for the Joint Operations reference. Back in the day I acted as spotter for the boat artillery. You could walk the rounds in on top of campy snipers. Fun times. Multi-day matches!
Toally agree they need some dedicated hardware to build up a core set of systems for multiplayer
Last edited:
There is no "right" because its an unsolvable problem. The OP was describing the CAP theorem which states that you cant have all three of consistency availability and partition tolerance and a real time game essentially requires all three. The client server model is the goto model because it doesn't distribute the consistency aspect - there is one master game world state. Essentially a P2P game is not a distributed system, it is a monolithic system with its components split over a network and hence fails whenever any peer fails. It also falls foul of Metcalfs law hence each additional peer results in a geometric increase in data exchange and geometric increase in the probability of data loss.
Last edited:
Exactly. (You don't happen to be looking for a job in or around Vienna, Austria perchance?)
Seriously though, I regret my choice of thread title and wish I could change it. It comes across as me going "booooh FD doesn't know what they're doing" - which is not the case. They knew why they chose P2P and I think it was the right choice - for the most part. Which is why I am absolutely not proposing to drop P2P. It makes perfect sense for such a large galaxy. And I think their work is commendable for a computer game. Some of the issues simply cannot be helped, like someone with a bad connection joining your instance, and dragging everyone else's experience down. Frontier did a great job with their matchmaking and island health checks.
I merely propose Client/Server for warzones, capital ship battles, and hand-picked locations. That way, you can not only put more players into a single warzone, but you could actually make capital ships something that exists only once in the game. Destroy the capital ship and it's gone and has to be rebuilt. No more "oh I'll just jump out and back in and it'll respawn again".
As for some people who said that this is impossible because of the amount of data ED transfers, this is simply not the case, and you don't even need wireshark to confirm this. The fact of the matter is that ship movement needs much less information to be transferred than, for example, the many hundred players in an MMOFPS like Planetside. They don't just move in a straight line doing slow turns like Elite ships, they tiptoe left and right, back and forth, they swivel around on the spot. Their changes in direction are much more rapid, hence more data to transfer.
I think C/S is a viable addition to Elite Dangerous, supplementing P2P at crucial hotspots and enabling this game to truly become a massive game.
Just imagine another Hutton Mug run, but this time with hundreds of commanders forming a giant convoy in supercruise, heading towards that station. (of course, you'll want P2P instances for the actual station arrival so people can realistically dock... but you get the idea)
Now, whether or not Frontier has the resources, or even the inclination to look into something like this depends on whether or not they believe that their current solution is "good enough".
I'd hate to see them shrug and declare things as good enough, don't you agree?
(And yeah, P2P matchmaking needs improvement too, as it is so very frustrating to not end up in the same instance despite being winged up)
Last edited:
All you can play offline are the tutorials, I wouldn't call that most of the Game
Problem with the limitation is not just P2P, a Warzone with 100 of Players is something that would require a whole lot of Data to go back and fort. Too much for your standart internetconnection.
world war 2 online (battleground europe) has a 5000 player server, half scale map of europe, fighting on land sea and air all going on at the same time with no instancing or map loading. and thats using unity engine. 100 player war is nothing
I wonder if perhaps a custom BGP-ish routing table that ranks the AS networks by latency based on peer-provided stats could be used with appropriate caching and replication to quickly to organise island/instance peers into some kind of cluster topology?
Authority (and responsibility for aggregating and forwarding hit and movement data) could be assigned to root nodes based on ratings for for latency, bandwidth, reliability (and client trustworthiness), with leaf nodes populated by the lesser-rated clients.
Total bandwidth usage could be reduced somewhat in larger instances, possibly allowing the player cap to be raised a bit - and the effects of bad peers could be mitigated somewhat. There's probably plenty of issues with updating nodes as players came and went, and I'm sure it's a bad idea from the off, but it seems like it might at least have some merit to my addled mind.
Authority (and responsibility for aggregating and forwarding hit and movement data) could be assigned to root nodes based on ratings for for latency, bandwidth, reliability (and client trustworthiness), with leaf nodes populated by the lesser-rated clients.
Total bandwidth usage could be reduced somewhat in larger instances, possibly allowing the player cap to be raised a bit - and the effects of bad peers could be mitigated somewhat. There's probably plenty of issues with updating nodes as players came and went, and I'm sure it's a bad idea from the off, but it seems like it might at least have some merit to my addled mind.
Note that WW2 online uses a custom engine (I believe it was based on Granny3D initially but I may be mistaken), which they call "Unity". This is not the same as the popular hobbyist engine with the same name
The way these games (Planetside2 too) work is that they form "bubbles" around the player as a kind of semi-instance. These bubbles are shrunk or enlarged depending on how many players are in a certain area. Usually this works great, it's just when there are really lots of players in a small area that the bubble size gets so small that enemies can just "pop in" in front of you. Planetside2 is particularly impressive in that it handles different units differently, i.e. tanks and antitank-turrets are sent and rendered from much farther away than foot soldiers who just have a rifle. An ingenious way to try and fix the issue of tanks getting shot by players not visible to them. It works really well and I strongly recommend everyone to give Planetside 2 a try sometime, it's really impressive. It's also one of the few F2P MMOs that are really perfectly enjoyable and fair without paying a dime, as weapons aund unlocks are mostly sidegrades and highly situational, not straight upgrades.
Lets not forget that the Networking is now far far far better than it was on day 1.
Often struggled to see more than 3/4 players in any instance, even at the start locations, now if go anywhere, especially places like CGs, you'll see plenty of people.
Often struggled to see more than 3/4 players in any instance, even at the start locations, now if go anywhere, especially places like CGs, you'll see plenty of people.
I consider ''wings'' and ''cqc'' non working features, because I have to work around them a lot (multiple re-logs to game/rejoins to the wing if others don't (or suddenly stop) see me or vice-versa, CQC finds me a match only if I wait at least 30 minutes - which is insane amount of time).
So yeah, I wouldn't mind having client-server instead of P2P in certain locations in game. Would be good to add them at current CG-goal systems and other POI mentioned in galnet (cap ship battles maybe as you say), new players spawn systems too and CQC matches. This would let devs add some scripted events without updating our clients first and spoiling secrets.
Oh that is true. Though... sadly (or luckily) you sometimes don't even notice when things just don't work. Sometimes you can jump into Eravate and there's nobody there for some weird reason - turns out matchmaking simply failed and you're essentially playing singleplayer, oblivious to dozens of other players whom are just invisible to you. You may think that the game is empty, but when you look into the network logs you see that the game tried to join an instance, failed, and then dropped you into an empty one instead.
Or worse, matchmaking doesn't even connect you to an instance with your wingmate in it because the island health is already pretty low, regardless of player count. Remember, all it takes is one player with a crappy connection and not only does the game become laggy, but it also prevents others from joining in. However I fear we'll have to live with that, at least for the P2P instances.
What would be REALLY cool of course would be that frontier could just spawn lots and lots of generic server instances which then could just be used as they are available instead of P2P sessions. You could still do dedicated servers for specific locations, but you could also have a pool of extra servers for any location in the universe. Once matchmaking determines that there's no free server instance available, it could always fall back to P2P.
Last edited:
Completely agree... The instancing was really bad in the early days, much better now.
As I understand it that's how the more recent CoD games did it: have a pool of dedicated servers, but just have it P2P-hosted if there aren't any available. It's a smart way of doing it, since it allows for transparently scaling the server capacity up/down based on various factors (performance, player count, budget).
apparently i have to give rep to someone else before i can give it to the OP. don't remember giving you some rep already.
anyways... i support this sentiment. client/server is really necessary. with peer2peer, i always see rubberbanding players in supercruise, if i see other players. i have a somewhat decent internet connection. 1.6mbit down, 128KB up. this should be enough.
it's a pity that the multiplayer performance is that bad. instead of improving it, we get an expansion that doesn't seem to address anything about network performance.
to this date, i don't even know how CQC runs. does it also use a peer2peer architecture?
i'm not questioning that peer2peer is future proof. but now, not all users have broadband internet at home. so... it's not an optimal solution.
anyways... i support this sentiment. client/server is really necessary. with peer2peer, i always see rubberbanding players in supercruise, if i see other players. i have a somewhat decent internet connection. 1.6mbit down, 128KB up. this should be enough.
it's a pity that the multiplayer performance is that bad. instead of improving it, we get an expansion that doesn't seem to address anything about network performance.
to this date, i don't even know how CQC runs. does it also use a peer2peer architecture?
i'm not questioning that peer2peer is future proof. but now, not all users have broadband internet at home. so... it's not an optimal solution.
Last edited:
Wouldn't they run out of hosted instances pretty quickly? The game creates instances for each system's super cruise and anywhere a player drops out. So even if you had a few thousand hosted instances the majority of players would still experience P2P while travelling around and I'm sure they wouldn't think the experience was any better overall.
Using a dedicated server for CQC makes sense, but its hard to imagine the amount of work & resources that would be required to support it in the main game.
- - - Updated - - -
The network code has improved with various updates. The problem is that there isn't a silver bullet solution. The idea that adding dedicated servers somewhere fixes everything is a bit of a fallacy.
- Status