I’m talking about applications in general, not just MMOs. Lots of kinds of companies used to have these problems. I don’t know of a reason that this problem would be an entirely different beast in an MMO than in any other kind of application where many users are interacting in real time. E.g., an ecommerce app needs to know, in real time, whether somebody else has bought the last Looney Toons toothbrush, a stock trading system needs to exchange all kinds of data between millions of clients on the millisecond timescale. Most of those sorts of applications used to have this problem.
But definitely I may be missing something. I agree that as MMOs go, EvE has huge fights, but is there a reason that would be different than other kinds of applications that have huge numbers of users and real time info?
Yeah. I mean, no judgment in that. EvE was released in 2003, and few companies were thinking about scalability back then the way they do now. But seems like pretty much everybody else (see note above) solved it quite a while ago. I don’t see why CCP couldn’t.
I guess because all that code needs proper untangling first.
I don’t know. We’ll see eventually.
I think the current performance numbers are quite impressive already and …
… the more players can fight in one spot …
… the more players will show up to fight.
Eve is different from an e-commerce platform because it has to process a bunch of non-linear equations. An e-commerce platform is just simple algebra.
It might be possible to rewrite the code to make solving the grid just a matter of solving a stack of linear equations, but it would be a huge project and would probably force CCP to change some of the game mechanics.
LOL that’s an interesting point. It’s like the problem with building more lanes. Everybody always complains that some key highway is only 3 lanes and thinks they should make it 4 to alleviate traffic. So then they make it 4, it becomes a more attractive commuting option, and hence more drivers take it, and it is immediately back to the same traffic jams.
I dunno, though. I’m a former developer. I defend CCP devs on here periodically. E.g., the other day I was arguing that people shouldn’t be so hard on them for re-introducing old bugs because regressions are actually way harder to avoid than they sound like they would be. Overall, I think they do a pretty good job, and don’t have a huge team. But this architecture where systems can’t span nodes… That feels awfully outdated… Reminds me of the early 2000s when most multi-user applications had that limitation. Companies were trying to buy more and more massive servers to accommodate all kinds of applications that couldn’t span multiple servers. But that isn’t really a thing anymore. And, as big as EvE battles get, that number of users and that volume of data is nothing compared to what many enterprise type applications have to deal with.
I’m not sure how the complexity of the processing changes whether it could be spread across multiple systems. And EvE is hardly at the forefront of real time applications that make complicated calculations. E.g., there are many AI applications that make split-second decisions that are a bazillion times more complicated and run on clusters. In fact, the complexity of the processing generally correlates to more parallel architectures- the more complicated the processing, the greater the need to span nodes.
They don’t need to do thousands of square roots per second …
… which are super slow …
… next to all the other stuff …
… and likely don’t have python around their native code.
You might know more about this than I do - I was under the impression that AI was mostly Boolean math and a few integers - exactly the sort of math that computers are most efficient at.
Solving 100,000 polynomial floating-point equations is always going to be slow.
Not to forget that that’s easily pipelined on the CPU for really high throughput.
You won’t achieve that high throughput with all the math involved for what’s EVE doing.
I think it has something to do with data localization, but … no idea.
I can think of simple ways of making it easy and efficient to spread code out to different machines,
so there must be a really good reason why they can’t do it.
I don’t disagree with that, but I think that’s a different problem. That’s why we have 1 second ticks which may be difficult to reduce below 1 second, but isn’t why we have TIDI when there are too many ships. Spanning nodes should help them solve lots of equations at once, but not necessarily help them solve individual equations faster.
… Ok, I made it like 1/4th the way through the scroll and I had to stop because my brain cells were crying out in pain.
Getting rid of one of the major weapon types of this game is a horribly bad idea. First it would change a very considerable amount of pve and pvp setups that CCP has worked hard to balance. Second, this would eliminate basically all supers and carriers, which have huge purpose in both pve and pvp. So they would change part of the foundation of how big fights play out which could cause who knows how many issues.
But don’t get me wrong, I hear what you are saying. Instead of having a 3d model for drones, maybe on potato mode make them 2d billboards. Have them still be selectable in space and in overview, but reduce their poly count from idk how many to like 1-4 polies. Sure, that might slightly help.
Another option would be to remove drones from commonly used ships for those massive fleet fights. Like eagles, and munnins. Neither of those ships are used for pve very often, they don’t get drone bonuses, so they will hardly be missed… Except when they would drop EC drones to moonwalk out of a bad spot. But hey, give them a +1 warp core str and make everyone mad when the enemy fleet can moonwalk out.
Your suggestion is pretty bad for the game, because it would make drones immune to any damage and they could not be destroyed anymore. Your reasoning “would negatively effect small gangs” is just a tip of the iceberg… You could use invulnerable fighters to nuke keepstars off the grid (just launch the damn things from extended ranges and watch them chew the keepstar in pieces…)
There are better (less crippling to game mechanics) ways to reduce server/client load, for example I posted this thread while ago with some suggestions about it: TIDI 2.0 (those would leave normal non TIDI gameplay 100% untouched and only affect these big lagfests…)
Perhaps it would be possible for the server to “group” drones in large scale fights so, for example, a flight of 5 drones from a ship gets handled as a single entity? One of the drones would get buffed to the damage / TP / speed / tracking of the whole flight and the others get ignored by the server?
I know it would be kind of a patchwork, but it would only kick in for fights with heavy TIDI. If the “lead drone” get’s destroyed, the whole flight would be lost. Should reduce the server load in big battles.
Fighters on the other hand are a different problem.
That does nothing to TIDI (its client server network load, not graphics related). You can already put your client to potato mode (and even super potato-less mode that was introduced recently.)
