The Technical Side of EVE Online
Here at EVE Online Fanfest 2008, members of the press were treated to a handful of special presentations from some of the development team at CCP – and some of their associates as well. Unlike the vast majority of MMOGs, which split their playerbase into multiple separated “shard” servers, EVE is unique for having all of its quarter-million strong population on one collective realm. Naturally, this results in technical hurdles for the team to overcome – it’s easy for a traditional MMOG to just buy a new server cluster and clone a new shard, but that sort of standard solution just wouldn’t work for CCP and EVE.
I won’t lie to you – the technical details were a bit over my head. I did nab some of the specifics here and there: they apparently decide what their next projects will be via rugby scrums, they’re programming in a new language developed by barracudas (or perhaps they’ve opened a new office in Cuba, I’m still unclear on the exact details of this one), and they apparently manage their server clusters by sticking swords and knives into them. Something like that, anyway. I’m a bit fuzzy on the technical industry terms, to be fair.
What was clear, though, were the results:
Since the launch of the Premium graphics update to EVE, the shift towards rendering and generating textures and particle effects via the GPU instead of feeding the calculations through the CPU has reduced the drain on the systems significantly. Through optimization, we were told, they’d reached the point where, in one of their tests that they showed us, they’d managed to run ten separate instances of the EVE client on one (admittedly top-end) machine. Some of their further goals included procedural generation of planet surfaces that would allow players to zoom in far beyond the range where today they’d see the map break down into pixelization, or the full dense particle rendering of the EVE universe’s gas giants.
The CCP team also talked about the optimizations they’d made server-side, and the upgrades in functionality since one year ago. In November 2007, the most pilots that could be in Jita (the main market hub in the EVE galaxy) before horrendous lag set in and their support staff queues were filled with requests to move one’s ship out of the area was approximately 600. Now, they’ve had 850 simultaneously without even a single such problem rearing its head. While in the most crowded mission zones, at peak hours lag could get to the point where combat and information modules would take 20 or 30 seconds to update – hardly playable by anyone. They announced proudly that they’d reduced that lag tenfold, to where it was now a delay of 2-3 seconds at worst. Furthermore, large-scale fleet combat was now possible and playable with a fleet size of 1000 ships, up from just 400 a year ago.
While the new Quantum Rise expansion wouldn’t feature a graphical upgrade like the jump between Classic to Premium, the one after that would contain a similar jump – after that, though, they hope to update the game in smaller, more progressive chunk.
Stay tuned for coverage from Fanfest – tomorrow, we’ll get a closer look at walking in stations, and the CCP team has kept hinting that there’s a large announcement in store for Saturday.