EVE Technology Lab

'unhackable'. The only computer that can't be hacked is one that's switched off, in a sealed concrete bunker, protected by the SAS. And only just. Things can just be /difficult/ to hack.


The problem with big fights in eve is single threaded processes. If my memory serves, mainframes tend to be good at running lots and lots of threads at the same time. Which isn't particularly useful for Eve. Work is being done to split some bits off onto other threads, but the core 'X shoots Y, Y shoots back' pretty much has to be a single thread. (proper multithreading is difficult. And not possible for certain kinds of problem)

In addition to Steve's comment, there are additional layers of abstraction required in order to get modern server tech to talk with IBM "iron"--converting from EIBCDIC to ASCII and the like. I'm a mainframe (and distributed) programmer myself, but in examining the way that EVE works, I don't see it being a plausible solution. The advantage of the mainframe/MVS is in processing large quantities of centralized data (all stored in, essentially, the same "place"), but when it comes to having to maintain all that data on an individual basis--for each client connected to the server--the small overhead that would normally be incurred is increased exponentially.

I can give you other excellent answers:

- the top EvE issue does not come from a OS / hardware limitation but by legacy architecture which originated a decade ago. Therefore putting more servers or more powerful OSes is not going to help. Proof: reinforced nodes still bend over despite being "dedicated" to just one system. Even an expecially optimized one like Jita, where the server does not even manage all the operations, but those are split on multiple computers (one for the markets, one for the ships sim and so on).

- I have greatly enjoyed working with some AS/400 and even 1 mainframe (whose "hard disk" weighted so many tons they had to reinforce the building to not crush under its weight! ). However powering 20,000 end users filling in corporate forms is not the same as computing more or less complex mathematical simulations.

What would help EvE the most would be a fully "snap in" redundant "cloud alike" architecture, with basic operations being split in small chunks (unlike the current vertical 1 CPU heavy weight processes) processed in a "lazy".
For a very cheap implementation, I envision EvE being run by something like Seti@Home where each player hosts a "micro-sim", a little shardlet of EvE with anonymous data coming in to be processed and coming out with the results.
Of course that'd immediately prompt security issues (old MMO motto: "never trust the clients") but those shardlets could be sandboxed on the central EvE shardlet gathering process at CCP's HQ. Only a minority would be enticed to try malicious attempts, as the data coming to their computer would be anonymized (just IDs), possibly encrypted and anyway regarding a random event in a random other system in EvE.

Yes, in principle it's true. In practice there are some very basic yet hard to solve obstacles and MMO makers have to let the clients send data. I.e. most MMOs have "maps" / terrains. It's very very expensive to continuously check the clients characters positions on those maps, in order to decide if they are slamming into a closed door, wall, or falling in water (and adapt the animations to those situations). It's also very bad looking to have characters that enter a wall just to be bounched out after half second just because some lag delayed the server deciding you slammed into a wall.

So they trust the clients sending their positional data, but with the same approach I proposed: a "sandbox" that performs sanity checks on such data against the now classic hacks: teleporting, flying, bypassing obstacles, flattening mountains (client side) and so on.