If YOLOL chips are runned client side. Wont everyone see something different in big faction factory cause multiple people will work on it?
- Thread starter TGess
- Start date
- Joined
- Aug 25, 2019
- Messages
- 67
As I understand the game's network architecture, whenever there's one or more players within render range of a construct, the client of one of those players is assigned as the construct's "technical owner" responsible for simulating that construct and sending updates on its state to other players and the persistence server. As such, there's only ever one authoritative state for a construct, though it is possible for clients rendering it to get desynchronized due to latency from each other.
If no players are in range, the construct goes dormant and its last known state is saved into server memory. In order for an automated factory to run 24/7, it either needs to be in a heavily trafficked area or somebody needs to AFK there.
If no players are in range, the construct goes dormant and its last known state is saved into server memory. In order for an automated factory to run 24/7, it either needs to be in a heavily trafficked area or somebody needs to AFK there.
There was some talk about "world anchors" that would keep pockets of space running on dedicated servers.
It was just thrown idea, not a plan or promise.
Personally I'd provide special client that can run (simulate) pocket of space of chosen size without rendering it.
So players can run factories on optimized dedicated servers, if they want.
Or even run it from ordinary PC, but thanks to no graphical rendering and possibly much smaller space it would be much lighter than normal client.
It was just thrown idea, not a plan or promise.
Personally I'd provide special client that can run (simulate) pocket of space of chosen size without rendering it.
So players can run factories on optimized dedicated servers, if they want.
Or even run it from ordinary PC, but thanks to no graphical rendering and possibly much smaller space it would be much lighter than normal client.
Plus even if you have one dedicated pc as a faction it might not be enought to cover whole factory. We dont know how far it goes + it could screew up whole timing of that factory if some parts start working earlier cause the load in earlier. So I hope yolol chips load in memory at once and start working at once. Plus in PVP if my pc is the one running all the simulations and I get low fps cause of that I am gonna be very unhappy.
I own a desktop and a laptop and use a separate office computer at work. My computer is sitting at home alone connected to the internet and online pretty much 24/7. My laptop goes where I go. My work computer lives at the office, but is also sonstantly connected to the internet and constantly on if it needs to be.
This is the situation almost everyone I know is in. They have one or more computers, and their home PC sits at home connected to the internet when they're not using it.
Idling AFK is effectively cost neutral and because of the distributed server model Starbase is using doesn't have the issues you see with it in traditional server-centric games. You going AFK to keep a node online is actually saving frozenbyte server resources and for most people, just using a PC that would be doing nothing to help run their space empire.
This is the situation almost everyone I know is in. They have one or more computers, and their home PC sits at home connected to the internet when they're not using it.
Idling AFK is effectively cost neutral and because of the distributed server model Starbase is using doesn't have the issues you see with it in traditional server-centric games. You going AFK to keep a node online is actually saving frozenbyte server resources and for most people, just using a PC that would be doing nothing to help run their space empire.
The world anchor was an in game item someone before proposed that would act link a chunk loader in minecraft. I wasn't criticising the idea of leaving a computer on but I don't think it's flawless. Managing things client side means that it's very hard to make sure players don't alter what they send to servers to give themselves an advantage. For instance minecraft used to have a client side inventory and basically allowed players to spawn in whatever they wanted. If you are flying past an enemy ship that was previously dormant then how are the game servers to know you arent simulating it properly if you tell it to open all doors or something. Those situations would require complex heuristics to stop and even they can't ever be perfect. If they can pull it off though without massive hacks then more power to them but I am skeptical.
Minecraft did to inventories what most online games do out of the box these days. it put everything not critical to sim performance on the server. In the same way you can easily automate checks for speedhacks you can easily automate checks against this kind of behavior by authenticating player locations. The difference between starbase and a traditional servercentric model is that Frozenbyte isn't actually running the simulation servers, but they're still running central item, world state, and character data servers, and like any traditional MMO when a sim talks to the storage system there's usually a validity check in there somewhere.
Its basically just a really slow version of what all client/server games already do. When you move forward in an online game, you're telling you client "do the thing." and it says to the server "hey I did a thing"
In most online games the server then says "lemme check this thing you did to make sure its legit" and then sends back your new position, inventory, etc. to the client. This is why every game isn't full of item dupes and speedhacks. Your auth server's responsible for knowing the rules of the game and being able to tell when an action looks fishy (for example, 1000000 tons of ore appearing in an inventory without leaving another inventory with no rocks nearby)
The simulation may not be running on the auth server but all the economy and player account data is. Warframe and several other MMO-lite games use exactly the same system. Player clients host every game of warframe, but DE's servers house and authenticate all of the persistant economic, storage, customization, and progression data. Its a pretty well known implementation at this point, though nobody has tried to tie it together to distribute the load of a single persistant universe before, but I get where frozenbyte is going with the idea because it makes a lot of sense.
The secret sauce of starbase will have to be to smoothly transition players between two client nodes in some form, perhaps by using FBs in house servers as a bridge. This is where you see most games that use this approach being weak, if the host drops out everything screeches to a halt as the game slowly sends a snapshot of the current game state, selects another client, and then puts everyone back in place on the newly selected simulation. My assumption here is that Frozenbyte's trying to use in house resources to make that transition smoother, essentially using high performance in house servers as a bridge or a distributed computing mechanism to solve that problem.
If it works well, its a pretty revolutionary trick for running MMOs on the cheap and I'd love to see a whitepaper on the specific implementation. Done well it could actually make the central server model obsolete. Done poorly it could be annoying to play the game but not so much its a dealbreaker. I'm sure everyone would deal with occasional world freezes in exchange for everything else starbase offers.
Last edited:
I wasn't aware they used P2P but imo it would make it so much harder to avoid hacking. You can't design a heuristic for determining the result of a program without simulating it yourself so at some point corners will have to be cut in how much work Starbases servers will do. Control at some point is in the hands of the players simulating the universe and that is where they can do as they wish.
That's not entirely true. You don't need to model a simulation to check its validity. Much of the time automated checks are used to provide leads to a small number of GMs, flagging them for direct observation. Once you ban the first round of accounts word gets around and that exploit tends to vanish pretty fast.
A simple example is a speedhack test. Lets say the auth server already periodically stores the location of players, but it only does so when those players log out or change zones. Because the server knows the rules of the game state that there is an absolute galactic speed limit, it also checks a max distance when it stores that location. If a player is found to have moved further between logins than the rules of the game would allow, they're flagged and a GM ticket is generated to monitor them next tiem they log in.
Automated reporting tools like this are the secret sauce of GM governance in virtually every MMO. GMs don't just hang out waiting to catch people hacking.
A more complex example:
Ore check I theorized in the previous post doesn't require simulating the game to do cheat detection. It requires things the authoritative server already knows, inventory, environmental state, location of player. It doesn't need to run simulation ticks to be an effective implementation, nor does it need to process physics or damage modeling. It doesn't even need to run that often. Running this sort of integrity check once every few minutes (usually on a low priority thread so it only runs when there's less stuff for that server to do) is a pretty common way for client-centric games that phone home to central account servers to keep their games from becoming playgrounds for hackers.
A simple example is a speedhack test. Lets say the auth server already periodically stores the location of players, but it only does so when those players log out or change zones. Because the server knows the rules of the game state that there is an absolute galactic speed limit, it also checks a max distance when it stores that location. If a player is found to have moved further between logins than the rules of the game would allow, they're flagged and a GM ticket is generated to monitor them next tiem they log in.
Automated reporting tools like this are the secret sauce of GM governance in virtually every MMO. GMs don't just hang out waiting to catch people hacking.
A more complex example:
Ore check I theorized in the previous post doesn't require simulating the game to do cheat detection. It requires things the authoritative server already knows, inventory, environmental state, location of player. It doesn't need to run simulation ticks to be an effective implementation, nor does it need to process physics or damage modeling. It doesn't even need to run that often. Running this sort of integrity check once every few minutes (usually on a low priority thread so it only runs when there's less stuff for that server to do) is a pretty common way for client-centric games that phone home to central account servers to keep their games from becoming playgrounds for hackers.
The elegant solution would be to simply let the players pay the IRL cost of a "world anchor" through the game, if there is one. People pay for dedicated servers in all sorts of games, and having to pay a tiny monthly fee to run a station 24/7 is nothing.
Having the developer handle the business end lowers the cost for the end user (they can do it far more efficiently in every way). Save the planet by not having everyone running alts off their home machines 24/7.
Having the developer handle the business end lowers the cost for the end user (they can do it far more efficiently in every way). Save the planet by not having everyone running alts off their home machines 24/7.