Assembly Hall

Today's events in HED GP are just another example in a long line of fights where the outcome is determined by lag and by CCP's continuing refusal to supply proper hardware on which to run these nodes. When will this stop? When is enough enough? When will CCP finally bring some new hardware to the table that will allow fleet battles to occur without one side or another having to deal with infinite warp tunnels preventing hardeners from being activated, lag causing dreads to cycle but not actually do anything, and these fleet fights getting determined by who got their first rather than which side shows up in force with the right ships?

I'm not one saying I have the answer. Like just about everyone here in this community I don't have behind the scenes access to see what pieces of hardware are being used to run these "reinforced nodes" but the fact that 100% server load and 10% tidi are becoming the norm for fleet battles shows that whatever CCP is doing about this is too little, too late.

This battle didn't pop up in some unknown place, this wasn't some stealth attack out of nowhere, this battle was telegraphed in advance so CCP should have been able to handle it.

So CCP, what is the problem? What is the bottleneck? You are in a unique position to have many thousands of vested individuals in the Eve Online community who do this stuff for a living. If Amazon can keep it's servers up and responsive during black friday and Cyber Monday, and laugh off hackers' attempts to DDOS them with the Low Orbit Ion Cannon, why can't CCP build a MegaNode (I'm guessing that the Jita Supernode is still inadequate for this task) that will actually be able to hang with this level of traffic or use? If there is a technical problem, then reach out to the community, talk to people who work for groups used to crunching endless amounts of data. I'd be willing to bet there is at least one person very familiar with the inner-workings of supercomputers who could provide some very positive insight right here in this community.

So CCP, instead of giving us vague replies when we complain like "We're working on it," please trust in the community that supports you and reach out to us, or at least give us some idea of what you're actually doing to solve a game breaking problem that we are facing on a regular basis.

There is no better hardware.

The problem with EVE isn't its hardware, it's the 2004 software that doesn't properly scale to 2014 hardware.






Source

It means that "they are working on it" and "you need to be patient." Changing the way an existing system operates (especially one as large and complex as EVE) without completely scrapping and starting from square one (which takes almost as much time and results in any further improvements on the old system being "wasted") is a pain in the ass. Fix one thing and another thing (which may not be related to the change in any way) WILL break.

tldr: anyone who tells you "that will be easy to code and implement" is lying through their teeth.

edit:

EVE code is "single threaded" which means that it can't really take advantage of multi-core systems. Part of the reason it is this way is because...
- multi-core CPUs did not really exist at the time the base architecture for EVE was created (circa 1998 - 2003).
- actions that players take cannot be randomly processed and spat out. They have to follow a clear line of progression based on who started what first. That means that even with multi-threading you cannot process and execute an action further down the que until the ones before have been executed. So if one action is taking longer than the others to process and execute, the rest of the actions behind it have to wait. That is where the "bottleneck" is.


One of their solutions (that they are working on) is to separate parts of the system (like the database) from the main server so there will less to process. They call it "Brain in a Box".

However this is no "silver bullet." Realistically it will just raise the threshold for when things get FUBAR.

Wrong.

There is such a thing like dynamically synced threads, whereby you create one thread per player, that can be processed in parallel and if players interact it's only then that they start waiting for each other as the threads inter-twine. Even then each thread is capable of consuming one core that means that 100 players CAN consume 100 cores thus paralleling the workload. Unlike now where all 100 players are stuck on 1 single core while the rest of the cores sit idle, mostly disabled to allow for that 1 single core to be overclocked to hell and back.

I believe EVE is in the state that SOE's engineers have called the overbloated EverQuest Code.

The legacy code, which would have to be completely re-written to allow certain changes to itself be processed and no amount of adding new stuff is going to fix that.