11:15 - 12:15
Liberating Live Service Crashes:
How 'bout a nice cup of STABILI-TEA?
Description:
Live service stability is hard because: No matter how good your QA is, they'll never catch as many issues as millions of players using a wide variety of hardware setups. When the game crashes you're not going to get much information, just a callstack in most cases with very little context. Logs are your best friend in development but logging is not quite as easy in production for security and optimization reasons. You do have to balance your engineering efforts between addressing bugs/crashes and producing new content to keep players happy and make enough money to sustain the studio. 50% of your crashes can be caused by 1% of your bugs, but these tend to be sneaky edge cases that are very hard to reproduce. This means that when your QA finds an issue during testing, it's hard to know if it's a 'one-off' that happens one in a million or if it's closer to 1 in 10.
If your game is procedural and systems driven like Helldivers 2, this makes it even harder because it's literally impossible to test every possible combination of weapons, stratagems, level-tiles, enemies, biomes, etc There are many tricks up our sleeves to solve or mitigate some of these issues. Learn in this fireside chat!